16 - 18 march 2015

11 downloads 4774385 Views 32MB Size Report
Mar 16, 2015 - The present publication, carrying about one hundred full-length research papers ... have been shortlisted after a careful review of more than two ...
Co-sponsors

978-93-82288-63-3 Isbn: Online Isbn:978-93-82288-54-1 Print

Proceedings of International Conference on Advances in Computers, Communication and Electronic Engineering

16 - 18 march 2015 Department of Electronics and Instrumentation Technology तमसो मा ज्योित

UN

IVE

RSIT

‫ﻟﻨﻮر‬ ‫ا‬

‫ﻣﻦ اﻟﻈﻠ ٰﻤﺖ‬ ‫اﻟﯽ‬

य गम

Y OF KAS

HM

IR

University of Kashmir, Srinagar, J & K

COMMUNE - 2015

University of Kashmir

Proceedings of 2015 International Conference on

Advances in Computers, Communication, and Electronic Engineering

ISBN (Online): 978-93-82288-63-3 ISBN (Print): 978-93-82288-54-1 Publisher: University of Kashmir, Hazratbal, Srinagar, 190 006,

J&K, India. Publication Date: 16 March, 2015 Editor: Dr. Mohammad Tariq Banday Copyright Notice: © All rights are reserved by the Department of Electronics and Instrumentation Technology, University of Kashmir, Hazratbal, Srinagar, 190 006, J&K, India.

University of Kashmir

Foreword The Department of Electronics and Instrumentation Technology had the privilege of organizing the 2015 International Conference on Advances in Computers, Communication, and Electronic Engineering (COMMUNE-2015), an erudite gathering of learned academicians, scholars, and students from across the globe. This is to acknowledge the determination and efforts made by the faculty, research scholars, students, and administrative staff of the department in organizing such a huge gathering that carried an aura of international scope. The idea to convene COMMUNE-2015 was conceived by the dedicated and young faculty members in the department who always stand up to the occasion for undertaking all sorts of academic and research activities. The title of the conference was selected in a way to ensure participation of academicians and researchers from a number of academic disciplines. To target the anticipated academicians and scholars globally, a dynamic website solely prepared for this purpose was created. The response we received across globe reflected the twin successes, both with regard to the selection of the theme of the conference and the wide publicity it received. Absolute professionalism was maintained in the selection, evaluation, and review process of the articles and research papers. The international academic standards were complemented by the local traditions of hospitality to provide the visiting delegates an ambience of camaraderie and warmth. As the head of the varsity, I certainly feel proud to be part of such an excellent academic activity. Being a first of its kind to be organized by the department of Electronics and Instrumentation Technology, in particular and University of Kashmir in general, COMMUNE2015 will go a long way in shaping the future policy of the Institution in organizing seminars and conferences. It was in this spirit that the organizers are determined to turn COMMUNE into an annual activity. The fruition of COMMUNE-2015 in the form of these proceedings has been prepared with diligence as per the standard norms in academic publishing. Great efforts were put in to prepare these proceedings in time to allow the visiting delegates to take these along at the culmination of the conference. Along with authors, the names of respected reviewers have also been included in the proceedings as a mark of compliment to their efforts. I hope these proceedings will surely serve an excellent reference book for scholars and researchers of various denominations around the globe. I wish all the participants a great time ahead.

Prof. Khurshid Iqbal Andrabi (Honourable Vice-Chancellor, University of Kashmir) Chief Patron, COMMUNE-2015

2015

iii

University of Kashmir

Foreword The 2015 International Conference on Advances in Computers, Communication, and Electronic Engineering (COMMUNE-2015) organized by the Department of Electronics and Instrumentation Technology from 16th to 18th March is a landmark in University of Kashmir’s tradition of bringing together academicians, scholars and learned men from across the world to discuss and deliberate on variegated issues of knowledge. With a sense of satisfaction, I acknowledge the efforts put in, and the enthusiasm shown by various stakeholders of the varsity to organize a conference of such a magnitude. The conference has been conceived with the idea to underscore the scientific information interchange between researchers, developers, engineers, students, and fellow citizens working in and around the world. Through presentations, special talks, panel discussions and networking, COMMUNE-2015 provided an excellent avenue for budding researchers and academicians to discuss with and learn from the established academic community in the field and infact served a motivation for scholars of various denominations to approach the problems at hand from an interdisciplinary perspective. In their pledge to make COMMUNE an annual affair, I promise to extend all kinds of support to the organizers. With a broader aim to promote research and developmental activities in Electronics, Computer Science, and Communication Engineering, I have been given to understand that the conference has attracted the attention of academicians, researchers and learned scholars from numerous institutions, research centres and universities within and outside country. Around 200 quality research papers addressing the pressing issues in the field of Electronics, Computer Science and Communication Engineering, were received by the organizers. With serious efforts, these were peer reviewed and evaluated based on originality, technical and/or research content/depth, correctness, relevance to conference, contributions, and readability. The papers presented and included in the proceedings cover multiple themes and ideas that are currently being addressed world over in the field of Electronics, Computer Sciences, and Communication Engineering. I believe that the proceedings will serve an excellent reference book for the scientific community and certainly will stimulate endeavors of further research.

Prof. Mohamad Ashraf Wani (Dean Academic Affairs, University of Kashmir) Patron, COMMUNE-2015

2015

v

University of Kashmir

Foreword With a sense of accomplishment, I write this foreword for the proceedings of the 2015 International Conference on Advances in Computers, Communication, and Electronic Engineering (COMMUNE-2015) organized by the Department of Electronics and Instrumentation Technology from 16th to 18th March. Seeking academic excellence has been the benchmark of University of Kashmir since its inception and concerted efforts are always being made to organize scholarly interactions in the form of seminars and conferences. COMMUNE2015 has been a right step in that direction and credit must go to its organizers whose hard work, sincerity, and perseverance made it a successful endeavor. The idea underlying COMMUNE-2015 has been to create a platform for academicians, researchers, and scientists from across the globe to present their work on the wide variety of modern day issues in the field of Electronics, Computer Sciences, and Communication Engineering. The presentations, discussions, and talks delivered during the conference were also aimed to motivate the budding research scholars and academicians on their threshold, to imbibe the scientific temperament displayed by men of knowledge present in the institution during these days. COMMUNE-2015 was successful in creating a community of reputed scientists and scholars, upcoming academicians, research scholars and students who have vowed to make it an annual activity. COMMUNE-2015 received wide publicity, especially through the websites exclusively dedicated to it and as such, the conference attracted scholars, scientists, and academicians from all over the globe. In addition, given the theme of the conference that encompasses a host of academic disciplines, as many as 239 research papers were received by the organizers. With absolute professionalism, the papers to be presented were assessed for their relevance to the conference theme, their novelty, and technical correctness, besides peer review. I congratulate the organizers for successfully culminating such a commendable task. I hope the present proceedings will definitely add new facts and details to the various domains of knowledge covered by it and will serve as a reference book globally.

Prof. M. Y. Shah Dean, Faculty of Applied Science and Technology University of Kashmir

2015

vii

University of Kashmir

Foreword The contemporary world is referred to as being in the "age of technology." Twenty years ago, the world had no Facebook, no Google, and no YouTube! There were no Smart-phones, no Bluetooth, and no Wireless Internet. Today, these things are such a part of our lives that we fail to realize that they have not always been there. Human-computer interface has advanced to such an extent that at times, it is difficult to do without it. Technology is as entrenched into our everyday life as never before. The economies of nations have become directly proportional to the advancement in technology, particularly to those pertaining to electronics, communication, and computers. A process of continuous exploration and innovation in these technologies is, thus, vital for being competitive in the rapidly changing technological world. The International Conference on Advances in Computers, Communication, and Electronic Engineering (COMMUNE-2015) held from 16th to 18th March, at University of Kashmir, is a significant effort in this direction. COMMUNE-2015 has provided a platform for researchers, academicians, and engineers to contribute towards exploring new trends and technologies. The objectives of the conference have been to create an avenue for the researchers, to present high-quality research and to be involved in professional interactions for the advancement in electronics, communication, and computers. The present publication, carrying about one hundred full-length research papers, is a valuable contribution as it addresses the most pertinent and upcoming advancement in computers, communication, and electronic engineering. The research articles embodied in this special issue have been shortlisted after a careful review of more than two hundred qualified submissions. I would like to thank all those who have been involved in the preparation of this manuscript, especially those who assisted in the review process. I deeply acknowledge and appreciate the efforts of the Organizer COMMUNE-2015 as well as those of the faculty, students, and the Research Scholars of the Department of Electronics and Instrumentation Technology, University of Kashmir, for their painstaking effort to organize COMMUNE-2015 within a short period of time.

Prof. G. Mohiuddin Bhat Dean Faculty of Engineering & Head of the Department Convener, COMMUNE-2015

2015

ix

University of Kashmir

Editor’s Note With the ever-growing interaction of technology with individual and collective lives in societies across the world, the challenges posed to technologists, scientists and academicians have multiplied manifold. This is especially true of Information and Communication Technology (ICT) that has now become the inevitable part of modern life. There is no denying to the fact that the life of 21st century human beings is inextricably intertwined with the varied manifestations of ICT. However, it is also certain that the ICT revolution that the world is presently witnessing has been successful because of the unprecedented advances made in Electronic Engineering. Advances in Electronic engineering over the years have subtly worked at the back end of this ICT boom. Developments in High performance VLSI devices and circuits, Monolithic Integrated circuits, Silicon Nanoelectronics, to name a few have efficiently contributed towards bringing in effectiveness in ICT. The broadened scope of ICT has now in its folds, issues and challenges concerning design of low power, high efficiency and small chip area VLSI circuits, bio-information, cryptography, information security, digital forensics, data hiding, artificial intelligence, etc. COMMUNE-2015 was organized with the intention of assembling scholars and academicians from around the world to deliberate on these issues concerning Electronics, Communication, and Computers in conformity with the current trends of interdisciplinary research. As such, the theme of the conference was purposefully chosen to ensure participation from scholars affiliated to a wide range of academic and research disciplines. The overwhelming response shown by scientists, academicians, and scholars from within and outside country from Computer Science, Electronics, Physics, Mathematics, Statistics, Linguistics, etc. reflect the diversified research areas subsumed under the theme. COMMUNE-2015 had eleven sessions including a virtual session, eight keynote addresses, and three expert lectures. The keynote address by Professor Chaturvedi described the characteristics of a cognitive radio as an intelligent device that can use side information about its environment to improve spectral utilization. Professor Sarkar in his keynote address shared hefty number of tricks and techniques to overcome current design challenges in building efficient and low power consuming VLSI circuits within minimum chip area. The third keynote speaker, Professor Chaudhury underscored the importance of deep neural network in building a hybrid text recognizer for document image analysis. Professor Ansari in his address highlighted various perspectives of IT ACT 2000 and deliberated upon management of IT security issues at technical and organizational levels. Dr. Kaushik discussed Spintronics based magneto-resistive memories in terms of their architecture, operation and compared them with conventional memories. In his second keynote, Dr. Kaushik highlighted the prospects of a promising interconnect material “Graphene nano-interconnects” and discussed challenges involved therein. Professor Lehal, reflected upon “the Sangam”, a machine transliteration system transforming Perso-Arabi to Hindi Script. Professor Beg, reflected upon advances in Mobile and Wireless Communication technologies such as coding and compression that could permit flawless telepresence wherein elements such as smell, touch, and taste can also be transmitted. Professor Bhat, while highlighting various Government initiatives such as STIP-2013, GIAN and PRIM for promoting innovation

2015

xi

2015 led entrepreneurship summarized the prospects of ‘Make in India’ initiative and challenges involved in its implementation. Mr. Rayees Mohammad Bhat highlighted emerging cybercrimes such as cyber terrorism, spamming, phishing, online theft, credit card frauds, etc. and discussed a few such cases reported from the State of Jammu and Kashmir. Professor A. H. Mir reflected upon checking authenticity of images using watermarking and stenography and detecting forgery through second order statistical approaches. The organizers received 239 full-length papers in total via Easy Chair submission system, which were reviewed by a team of 141 international and national reviewers. Based on the review, 98 papers were selected for presentation in the COMMUNE-2015. Apart from the review process, all the selected papers were strictly checked for plagiarism using the ‘Turnitin’ software and the authors were given the feedback as well. As per international academic standards, some authors were asked to drop the level of quoted work to below 15 percent of the total word count of the camera-ready paper. For as many as 90 selected papers, the COMMUNE-2015 received registrations, camera-ready copies, and online copyright transfers. Camera Ready copy of each paper submitted to the COMMUNE-2015 was checked for formatting errors and were corrected by organizing committee members. To publish the proceedings of a conference of COMMUNE-2015’s magnitude, it takes considerable amount of time and energy. However, by the grace of Almighty Allah and the perseverant efforts and dedication of the team behind COMMUNE-2015, presenters were handed over a copy of the proceedings at the inaugural function. It was indeed a moment of pleasure for the members of organizing committee to have accomplished such an uphill task. The seeds of COMMUNE were sown quite early in 2011, when the idea of convening such a conference was conceived by the current organizers under the chair of late Prof. N. A. Shah (May Allah bless his soul) and COMMUNE-2015 represents its first flower. The future will see many such colourful flowers as it has been decided to organize COMMUNE annually, in sha Allah. As member of the organizing committee and the editor of the proceedings, I would request your cooperation and seek your suggestions in making the COMMUNE more successful in future. Lastly, I would like to express my gratitude to Prof. Khurshid Iqbal Andrabi, the Honourable Vice-Chancellor, Prof. Mohamad Ashraf Wani, Dean Academic Affairs, Prof. Sheikh Javeed Ahmad, Dean Research, Prof. M. Y. Shah, Dean, Faculty of Applied Science and Technology, and Prof. G. Mohiuddin Bhat, Dean Engineering and Head of the Department for their outright support and patronage.

Dr. Mohammad Tariq Banday Editor Department of Electronics and Instrumentation Technology University of Kashmir

2015

Organization Chief Patron: Prof. Khurshid Iqbal Andrabi, Honourable Vice-Chancellor, University of Kashmir. Patron: Prof. Mohamad Ashraf, Dean Academic Affairs, University of Kashmir. Patron: Prof. Sheikh Javeed Ahmad, Dean Research, University of Kashmir. Convener: Prof. G. M. Bhat, Dean Engineering and Head, Department of Electronics and Instrumentation Technology, University of Kashmir. Organizer: Dr. M. Tariq Banday, Sr. Assistant Professor (Coordinator UGC-SAP), Department of Electronics and Instrumentation Technology, University of Kashmir. Advisory Committee: Prof. Khurshid Iqbal Andrabi, Honourable Vice-Chancellor, University of Kashmir. Prof. Mohamad Ashraf, Dean Academic Affairs, University of Kashmir. Prof. Sheikh Javeed Ahmad, Dean Research, University of Kashmir. Prof. Zaffar Ahmed Reshi, Registrar, University of Kashmir. Prof. M. Y. Shah, Dean, Faculty of Applied Science and Technology, University of Kashmir. Prof. Nisar Ahmad Rather, Dean, Physical and Material Sciences, University of Kashmir. Prof. G. M. Bhat, Dean Engineering and Head, Department of Electronics & Instrumentation Technology, University of Kashmir. Prof. S. M. K. Qaudri, Head, Department of Computer Applications and Director, IT&SS, University of Kashmir. Prof. Sharief-ud-din Pirzada, Head, Department of Mathematics, University of Kashmir. Prof. Manzoor Ahmad Malik, Head, Department of Physics, University of Kashmir. Dr. M. A. K. Baigh, Head, Department of Statistics, University of Kashmir. Prof. Ajaz Hussain Mir, Professor, Department of Electronics and Communication Engineering, NIT, Srinagar. Er. A. H. Moon, Director, NIELIT, Srinagar. Dr. Mohammad Ahsan Chesti, Department of Computer Science and Engineering, NIT, Srinagar. Dr. M. Tariq Banday, Sr. Assistant Professor (Coordinator UGC-SAP), Department of Electronics & Instrumentation Technology, University of Kashmir. Prof. Aadil Amin Kak, Department of Linguistics, University of Kashmir. Dr. Basharat Ahmad Want, Associate Professor, Department of Physics, University of Kashmir. Dr. Majid Zaman Baba, Scientist B, Directorate of IT & IS, University of Kashmir. Dr. Farooq Ahmad Khanday, Assistant Professor, Department of Electronics & Instrumentation Technology, University of Kashmir. Dr. Shabir Ahmad Parah, Assistant Professor, Department of Electronics & Instrumentation Technology, University of Kashmir. Dr. Javaid Ahmad Sheikh, Assistant Professor, Department of Electronics & Instrumentation Technology, University of Kashmir.

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

xiii

2015 Dr. Tariq Rashid Jan, Sr. Assistant Professor, Department of Statistics, University of Kashmir. Dr. Musavir Ahmad, Sr. Assistant Professor, Department of Linguistics, University of Kashmir. Dr. Nadeem Ahmad, Sr. Assistant Professor, Department of Library Sciences, University of Kashmir. Er. Abdul Mueed Hafiz, Assistant Professor, Department of Electronics & Instrumentation Technology, University of Kashmir. Er. Rouf ul Alam Bhat, Assistant Professor, Department of Electronics & Instrumentation Technology, University of Kashmir. Mrs. Farhat Roohi, Electronic Engineer, Department of Electronics & Instrumentation Technology, University of Kashmir.

Members of Organizing Committees: Dr. M. Tariq Banday, (Coordinator UGC-SAP), (General and Program Chair) Dr. Farooq Ahmad Khanday, (Finance Chair) Dr. Shabir Ahmad Parah, (Publicity Chair) Dr. Javaid Ahmad Sheikh, (Hospitality Chair) Mrs. Farhat Roohi, (Registration Chair) Er. Rouf ul Alam Bhat, (Accommodation Chair) Er. Abdul Mueed Hafiz, (Accommodation Chair) Mr. Nisar Ahmad Paray Mrs. Muzamil Hassan Mr. Azad Ahmad Shah Mr. Mohamad Rafiq Beigh Ms. Shafiya Afzal Sheikh Mr. Javeed Iqbal Reshi Mr. Nasir Ali Kant Mr. Mohammad Rafiq Dar Ms. Uzma Ms. Tawheed Jan Ms. S. Umira R. Qadri Mr. Farooq Aadil Rather Mr. Reyaz Ahmad Mathangi Mr. Jahangir Ahmad Mr. Mehboob ul Amin Mr. Imran Nazir Beigh Ms. Sakeena Akhtar Ms. Asma Nazir Naqash Ms. Jaipreet Kour Wazir Ms. Farhana Ahad Mr. Nazir Ahmad Mr. Zubair Ahmad Bangi

Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar

xiv

2015 Technical Program Committee Members: Prof. Aadil Amin Kak, Department of Linguistics, University of Kashmir, Srinagar, India. Prof. Abdul Quaiyum Ansari, Department of Electrical Engineering, Jamia Millia Islamia, New Delhi, India. Prof. Aijaz Ahmad, Department of Electrical Engineering, National Institute of Technology, Srinagar, India. Prof. Ajaz Hussain Mir, Department of Electronics & Communication Engineering, National Institute of Technology, Srinagar, India. Prof. Alam Aftab, Department of Physics, Indian Institute of Technology, Mumbai, India. Prof. Anurekha Sharma, Department of Electronic Science, Kurukshartra University, Kurukshartra, India. Prof. Anwar Shahzad Siddiqui, Department of Electrical Engineering, Jamia Millia Islamia (Central University), New Delhi, India. Prof. Carlos Molina, Department of Computer Science, Universidad De Jaen, Spain. Prof. Costas Psychalinos, Department of Electronics & Computers, University of Patras, Rio, Patras, Greece. Prof. Satya Prakash Ghrera, Department of Computer Science and Engineering, Jaypee University of Information Technology, Waknaghat, India. Prof. Ekram Khan, Department of Electronics & Communication Engineering, Aligarh Muslim University, Aligarh, India. Prof. Farooq A. Mir, Department of Law, University of Kashmir, Srinagar, India. Prof. Florina Ungureanu, Faculty of Automatic Control & Computer Engineering, Technical University of IASI, IASI, Romania. Prof. G. Mohiuddin Bhat, Department of Electronics & Instrumentation Technology, University of Kashmir, Srinagar, India. Prof. Mehraj Ud Din Mir, Central University of Kashmir, Srinagar, India. Prof. M. Mustafa, Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar, India. Prof. M. Salim Beg, Department of Electronics Engineering, Zakir Hussain College of Engineering & Technology, Aligarh Muslim University, Aligarh, India. Prof. M. Shah Alam, Department of Electronics Engineering, Aligarh Muslim University, Aligarh, India. Prof. Mainuddin, Department of Electronics & Communication Engineering, Jamia Millia Islamia, New Delhi, India. Prof. Mairaj-ud-din, Department of Electrical Engineering, National Institute of Technology, Srinagar, India. Prof. Mridula Gupta, Department of Electronics Science, South Campus, University of Delhi, Delhi, India. Prof. Najeeb-ud-din, Department of Electronics & Communication Engineering, National Institute of Technology, Srinagar, India. Prof. Nasib Singh Gill, Department of Computer Science & Applications., M. D. University, Rohtak, Haryana, India. Prof. Nilanjan Dey, Department of Computer Science Engineering, BCET, Durgapur, West Bengal, India.

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

xv

2015 Prof. Paresh V. Virparia, Department of Computer Science, Sardar Patel University, Vallabh Vidyanagar, Gujarat, India. Prof. R. K. Sarin, Department of Electronics & Communication Engineering, National Institute of Technology, Jalandhar, India. Prof. Roohie Naaz Mir, Department of Computer Science Engineering, National Institute of Technology, Srinagar, India. Prof. S. P. Singh, Electronics Engineering, Indian Institute of Technology, Roorkee, India. Prof. Santanu Choudhury, Department of Electronics Engineering, Indian Institute of Technology, Delhi, India. Prof. Seifedine Kadry, Department of Applied Mathematics, American University of the Middle East, Egaila, Kuwait. Prof. Shamim Ahmad Lone, Department of Electrical Engineering, National Institute of Technology, Srinagar, India. Prof. Sharief-ud-din Pirzada, Department of Mathematics, University of Kashmir, Srinagar, India. Prof. S. Naseem Ahmad, Department of Electronics & communication Engineering, Jamia Millia Islamia, New Delhi, India. Prof. Stefan Segla, Department of Mechanical Engineering, Technical University of Kosice, Kosice, Slovakia. Prof. Subir Kumar Sarkar, Department of Electronics & Telecommunication Engineering, Jadavpur University, Kolkata, India. Prof. B. A. Usha, Department of Computer Science Engineering, R. V. College of Engineering, Bengaluru, Karnataka, India. Prof. Vivek Kshirsagar, Department of Computer Science, Govt. Engineering College, Aurangabad, India. Prof. Yudong Zhang, School of Information Science & Technology, Nanjing Normal University, China. Dr. A. K. Daniel, Department of Computer Science & Engineering, M. M. M. University of Technology, Gorakhpur, U.P., India. Dr. Amit Kant Pandit, Department of Electronics & Communication Engineering, Shri Mata Vaishno Devi University, Katra, Jammu & Kashmir, India. Dr. Anwar Sadar, Department of Electronic Engineering, Aligarh Muslim University, Aligarh, India. Dr. Arman Rasool Faridi, Department of Computer Science, Aligarh Muslim University, Aligarh, India. Dr. Athar Ali Moinuddin, Department of Electronic Engineering, Aligarh Muslim University, Aligarh, India. Dr. Atul M. Gonsai, Department of Computer Science, Saurashtra University, Rajkot, India. Dr. Basharat Want, Department of Physics, University of Kashmir, Srinagar, India. Dr. Bharati Harsoor, Department of Information Science & Engineering, Poojya Doddappa Appa College of Engineering, Gulbarga, Karnataka, India. Dr. Brajesh Kumar Kaushik, Department of Electronics & Communication Engineering, Indian Institute of Technology, Roorkee, India. Dr. B. Sharada, Department of Computer Science, University of Mysore, Mysore, India.

Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar

xvi

2015 Dr. Carkis A. De La Cruz Blas, Department of Electric & Electronical Engineering, Public University of Navarre, Pamplona, Spain. Dr. Dharam Veer Sharma, Department of Computer Science, Punjabi University, Patiala, India. Dr. Dharmender Singh Kushwaha, Department of Computer Science & Engineering, Motilal Nehru National Institute of Technology, Allahabad, India. Dr. Faisal Anwer, Department of Computer Science, Aligarh Muslim University, Aligarh, India. Dr. Farooq Ahmad Khanday, Department of Electronics & Instrumentation Technology, University of Kashmir, India. Dr. Giuseppe Serra, School of Computer Science, University of Modena & Reggio, Emillia, Italy. Dr. Gurpreet Singh Lehal, Department of Computer Science, Punjabi University, Patiala, India. Dr. Harsupreet Kaur, Department of Electronic Science, University of Delhi, New Delhi, India. Dr. Hemraj Saini, Department of Computer Science & Engineering, Jaypee University of Information Technology, Wakanaghat, Solan-173234, India. Dr. Hung-Wei Chen, Prime Electronics & Satellites Inc., Taiyuan, Taiwan. Dr. Javaid Ahmad Sheikh, Department of Electronics & Instrumentation Technology, University of Kashmir, India. Dr. Javier Rubio Loyola, Centre of Research & advanced studies, National Polytechnic Institute of Mexico, Cinvestav, Mexico. Dr. Jitendra Agrawal, Department of Computer Science & Engineering, Rajiv Gandhi Proudyogiki Vishwavidyalaya, Bhopal, MP, India. Dr. Kandarpa Kumar Sarma, Department of Electronics & Communication Technology, Gauhati University, Guwahati, Assam, India. Dr. Kandarpa Kumar Sharma, Department of Electronics & Communication Technology, Gauhati University, Guwahati, Assam, India. Dr. Larrey Wen, Institute of Integrated & Intelligent Systems, Griffith University, Southeastern Queensland, Australia. Dr. Lian Wen, School of Information & Communication Technology, Griffith University, Brisbane, Australia. Dr. M. Hanumanthappa, Department of Computer Science, Bangalore University, Bangalore, India. Dr. M. Tariq Banday, Department of Electronics & Instrumentation Technology, University of Kashmir, India. Dr. Majid Zaman Baba, Directorate of IT & SS, University of Kashmir, Srinagar, India. Dr. Malaya Kumar Nath, Department of Electronics & Communication Engineering, National Institute of Technology, Puducherry, Karaikal, India. Dr. Mansaf Alam, Department of Computer Science, Jamia Millia Islamia, New Delhi, India. Dr. Manzoor A. Malik, Department of Physics, University of Kashmir, Srinagar, India. Dr. Mohammad Ahsan Chishti, Department of Computer Science & Engineering, National Institute of Technology, Srinagar, India.

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

xvii

2015 Dr. Mohammad Jawaid Siddiqui, Department of Electronics Engineering, College of Engineering, Aligarh Muslim University, Aligarh, India. Dr. Mohammad Nayeem Teli, Department of Computer Science & Engineering, National Institute of Technology, Srinagar, India. Dr. Mohammad Sarosh Umar, Department of Computer Engineering, Aligarh Muslim University, Aligarh, India. Dr. Mohammad Shukri Salman, Department of Electrical & Electronics Engineering, Mevlana University, Konya, Turkey. Dr. Mohammad Amjad, Department of Computer Engineering, Jamia Millia Islamia (Central University), New Delhi, India. Dr. Monica Mehrotra, Department of Computer Science, Jamia Millia Islamia, New Delhi, India. Dr. Musavir Ahmed, Department of Linguistics, University of Kashmir, Srinagar, India. Dr. Muheet Ahmed Bhat, Department of Computer Sciences, University of Kashmir, Srinagar, India. Dr. Musheer Ahmad, Department of Computer Engineering, Jamia Millia Islamia, India. Dr. Musiur Raza Abidi, Department of Electronics Engineering, Aligarh Muslim University, Aligarh, India. Dr. Navneet Agrawal, Department of Electronics & Communication Engineering, Maharana Pratap University of Ag. & Technology, Udaipur, Rajasthan, India. Dr. Nayan M. Kakoty, Department of Electronics & Communication Engineering, School of Engineering, Tezpur University, Tezpur, India. Dr. Norbert Herencsar, Department of Telecommunications, Brno University of Technology, Brno, Czech Republic. Dr. Nusrat Parveen, Department of Electronics, Islamia College of Science and Commerce, Srinagar, India. Dr. Omar Farooq, Department of Electronics Engineering, Aligarh Muslim University, Aligarh, India. Dr. Prashant M. Dolia, Department of Computer Science & Applications, Maharaja Krishna-Kumar-Sinhji Bhavnagar University, Bhavnagar, Gujarat, India. Dr. P. Thimmaiah, Institute of Technology, Sri Krishnadevaraya University, Anantapur, Andhrapradesh, India. Dr. R. D. Morena, Department of Computer Science, Veer Narmad South Gujarat University, Surat, India. Dr. Rafiqul Zaman Khan, Department of Computer Science, Aligarh Muslim University, Aligarh, India. Dr. Rahmat Widia Sembiring, CEO of Data Mining Research & its Application, University of Malaysia Pahang, Politeknik Negeri Medan, Indonesia. Dr. Raj Senani, Ex-Director, Netaji Subhas Institute of Technology, New Delhi, India. Dr. Rajeev Agrawal, School of Technology, North Carolina A&T State University, Greensboro, USA. Dr. Rajni Mohana, Department of Computer Science & Engineering, Jaypee University of Information Technology, Waknaghat, Solan, India. Dr. Rakesh K. Jha, School of Electronics & Communication Engineering, Shri Mata Vishnu Devi University, Katra, Jammu, India.

Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar

xviii

2015 Dr. Rakesh Vaid, Department of Physics & Electronics, University of Jammu, Jammu, India. Dr. Roshan G. Ragel, Department of Computer Engineering, University of Peradeniya, Peradeniya, SriLanka. Dr. Sajad A. Loan, Department of Electronics & Communication Engineering, Jamia Millia Islamia, New Delhi, India. Dr. Sanjay Jamwal, Department of Computer Science, Baba Ghulam Shah Badshah University, Rajouri, Jammu & Kashmir, India. Dr. Sanjay Tyagi, Department of Computer Science & Applications, Kurukshetra University, Kurukshetra, Haryana, India. Dr. Sanjeev Singh, Institute of Informatics & Communication, University of Delhi, South Campus, New Delhi, India. Dr. Sara Kadry, Software Research Institute, Athlone Institute of Technology, Republic of Ireland. Dr. Shabir A. Parah, Department of Electronics & Instrumentation Technology, University of Kashmir, Srinagar, India. Dr. Shikha Agrawal, Department of Computer Science & Engineering, Rajiv Gandhi Proudyogiki Vishwavidyalaya, Bhopal, MP, India. Dr. Shruti Jain, Department of Electronics & Communication Engineering, Jaypee University of Information Technology, Solan, Himachal Pradesh, India. Dr. Soumik Roy, Department of Electronics & Communication Engineering, Tezpur University, Tezpur, India. Dr. Subhash Chander Dubey, Department of Electronics & Communication Engineering, Govt. College of Engineering & Technology, Jammu, India. Dr. Sunil Kumar Wanchoo, Department of Physics, Shri Mata Vaishno Devi University, Katra, Jammu & Kashmir, India. Dr. Suresh Kumar, Department of Electronic Science, Kurukshartra University, Kurukshartra, India. Dr. Susheel Sharma, Department of Physics & Electronics, University of Jammu, Jammu, India. Dr. Svetlana Vasileva, International University College, Dobrich, Bulgaria. Dr. Syed Zaffer Iqbal, Department of Physics, Government College for Women Nawakadal, Srinagar, India. Dr. T. V. Prasad, Department of Computer Science & Engineering, Chirala Engineering College, Chirala, AP, India. Dr. T. Veera Kumar, Department of Electronics & Communication Engineering, National Institute of Technology, Goa, India. Dr. Tariq Rashid Jan, Department of Statistics, University of Kashmir, Srinagar, Kashmir, India. Dr. Tasleem Arif, Department of Information Technology, Baba Ghulam Shah Badshah University Rajouri, Jammu & Kashmir, India. Dr. Thomas Schlechter, Hardware R & D, Skidata AG, Groedig/Salzburg, Austria. Dr. Utpal S. Joshi, Department of Physics & Electronics, Gujarat University, Gujarat, India. Dr. Varsha Sharma, School of Information Technology, Rajiv Gandhi Proudyogiki Vishwavidyalaya, Bhopal, India. Dr. Vasudha Bhatnagar, Department of Computer Science, University of Delhi, Delhi, India.

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

xix

2015 Dr. Vijay Kumar, Department of Information Technology, Govt. of India, New Delhi, India. Dr. Virender Singh Kunda, Department of Electronic Science, Kurukshartra University, India. Dr. Vishal Goyal, Department of Computer Science, Punjabi University, Patiala, India. Dr. Vivek Chalotra, Department of Physics & Electronics, University of Jammu, Jammu, India. Dr. Werner Hartmann, Siemenns AG, Corporate Technology Research & Technology Centre, PET, Guenther-Schorowsky-Str, Germany. Dr. Zhi-Kai Huang, Nanchang Institute of Technology, Nanchang, China. Dr. Zhong Lin Wang, Department of Electronics & Communication Engineering, Georgia Institute of Technology, Atlanta, GA, USA. Er. Rajandeep Singh, Department of Electronics & Communication Engineering, Guru Nanak Dev University, Regional Campus, Sultanpur, Lodhi, India. Ms. Lafifa Jamal, Department of Computer Science & Engineering, University of Dhaka, Dhaka, Bangladesh. Ms. Maria B. Line, Department of Telematics, Norwegian University of Science & Technology, Trondheim, Norway. Ms. Pragya Dwivedi, Department of Computer Science & Engineering, Motilal Nehru National Institute of Technology, Allahabad, India. Mr. Jatinder Manhas, Department of Computer Science, University of Jammu, Jammu, India. Mr. Asad Mohammed Khan, Department of Computer Engineering, Aligarh Muslim University, Aligarh, India. Mr. Diana Palsetia, Department of Electrical Engineering & Computer Science, Northwestern University, Evanston, USA. Mr. Padma Prasada, Department of Electronics & Communication Engineering, Mangalore Institute of Technology & Engineering, Karnataka, India. Mr. Shibaji Mukherjee, Senior Manager at Oracle, Kinsight Analytics, Bangalore, India. Mr. Suhel Mustajab, Department of Computer Science, Aligarh Muslim University, Aligarh, India. Mr. Suryadip Chakraborty, Department of Computer Science & Engineering, University of Cincinnati, Cincinnati, OH - 45220, USA. Mrs. B. Shanmuga Priya, Department of Computer Science, Sri Ramakrishna College of Arts & Science for Women, Coimbatore, Tamil Nadu, India. Mrs. Shraddha Arya, Department of Computer Science, Sri Guru Gobind Singh College, Chandigarh, India.

Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar

xx

2015

Keynotes and Expert Lectures 

Information Theoretic Perspective on Cognitive Radio Networks [1] Prof. A. K. Chaturvedi, IIT Kanpur.



Design Challenges for Low Power VLSI Circuits [3] Prof. Subir Kumar Sarkar, Department of Electronics and Telecommunication Engineering, Jadavpur University, Kolkata.



Deep Learning for Document Image Analysis [5] Prof. Santanu Chaudhury, Department of Electrical Engineering, IIT, Delhi.



Security Issues in IT Systems and their Management [7] Prof. Abdul Quaiyum Ansari, Department of Electrical Engineering, Jamia, Millia, Islamia, New Delhi.



Spin Transfer Torque based Magneto-resistive Memories [9] Dr. Brajesh Kumar Kaushik, Department of Electronics and Communication Engineering, IIT, Roorkee.



Graphene Based On-chip Interconnects and TSVs: Prospects and Challenges [11]

Dr. Brajesh Kumar Kaushik, Department of Electronics and Communication Engineering, IIT, Roorkee. 

A Perso-Arabic to Indic Script Machine Transliteration Model [13] Dr. Prof. Gurpreet Singh Lehal, Department of Computer Science, Punjabi University, Patiala, India.



Moving Towards “Flawless Telepresence” Systems of the Future [15] Prof. M. Salim Beg, Department of Electronics Engineering, AMU, Aligarh.



Technology Innovation and Diffusion Practical approach towards 'Make in India' [17] Prof. G. Mohiuddin Bhat, University of Kashmir, Srinagar.



Current State of Cyber Crimes in the State of Jammu and Kashmir [19] Mr. Rayees Mohammad Bhat, IPS, SP, Hazratbal, Srinagar.



Detecting Forgery in Images: A Statistical Perspective [21] Prof. Ajaz Hussain Mir, NIT, Srinagar.

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

xxi

2015

Keynote - 1 Information Theoretic Perspective on Cognitive Radio Networks Prof. A. K. Chaturvedi* Indian Institute of Technology Kanpur, India

Keynote Cognitive radios hold tremendous promise to increase the spectral efficiency of wireless systems. We will start with a brief introduction to information theory and cognitive radio networks and then discuss the fundamental capacity limits in such networks. We will characterize a cognitive radio as an intelligent device that can use side information about its environment to improve spectral utilization. © 2015 Published by University of Kashmir, Srinagar. Selection and/or peerreview under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords: Cognitive Radio; Capacity; Degrees of Freedom; Interweave; Overlay; Underlay

*Speaker. Tel.: +91 512 2597613. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-001

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

1

2015

Keynote - 2 Design Challenges for Low Power VLSI Circuits Prof. Subir Kumar Sarkar* Department of Electronics and Telecommunication Engineering, Jadavpur University, Kolkata-700 032, India

Keynote Tremendous growth of the VLSI technology has been mainly due to progress of fabrication technology, which allowed systematic scale-down of device feature sizes and exponential growth of the integration level. Continuous device performance improvement is possible probably through a combination of device scaling, new device structure and material property improvement. Due to its small size, their potential integration level is significantly high and its low power operation solves some of the instability and reliability problems. The major challenges for design Engineers are to design new generation products, which consume minimum power, without compromising its performance or achieving minimum chip area. As we approach millennium, power dissipation has become the main design concern in many applications such as wristwatch, laptop, computers, and pace makers although early VLSI design did not consider it. The objective of such applications is minimum power for maximum battery life. Power dissipation is the greatest obstacle for Moore’s law. Modern chips consume power of which about 20% is wasted in leakage through the transistor gates. The traditional means of coping with increased power per generation has been to scale down the operating voltage of the chip but voltages are reaching limits due to thermal fluctuation effects. To save power, several tricks have been considered viz., minimize activity, glitches, effective capacitance, wire length of nodes and use of minimum possible supply voltage constrained by performance needed, design for high speed and then reduce voltage to get the desired speed. © 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords: Low Power; VLSI Circuits; Power Dissipiation; Tricks to Save Power; Glitches; Effective Capacitance

*Speaker. Tel.: +91 3324 572810. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-002 Delivered in Joint Session of SEEDS-2015 and COMMUNE-2015

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

3

2015

Keynote – 3 Deep Learning for Document Image Analysis Prof. Santanu Chaudhury* Dhananjay Chair Professor, FNAE, FNASc, FIAPR, Department of Electrical Enginering, I.I.T, Delhi

Keynote Deep networks provide a new paradigm for feature discovery and recognition. We can approach problems of document image analysis in the framework of deep learning. We shall examine use of deep learning for scene text recognition. Next we shall present an architecture for text recognition using deep LSTM. Text recognition involves some initial image processing steps like segmentation of lines and words which can induce error to the recognition system. Without segmentation, learning very long range context is difficult and becomes computationally intractable. Therefore, alternative soft decisions are needed at the pre-processing level. This paper proposes a hybrid text recognizer using a deep recurrent neural network with multiple layers of abstraction and long range context along with a language model to verify the performance of the deep neural network. In this paper we construct a multi-hypotheses tree architecture with candidate segments of line sequences from different segmentation algorithms at its different branches. The deep neural network is trained on perfectly segmented data and tests each of the candidate segments, generating unicode sequences. In the verification step, these unicode sequences are validated using a sub-string match with the language model and best first search is used to find the best possible combination of alternative hypothesis from the tree structure. Thus the verification framework using language models eliminates wrong segmentation outputs and filters recognition errors. © 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords: Feature Discovery; Document Image Analysis; Hybrid Text Recognizer; Deep Neural Network

*Speaker. Tel.: +91 9891 266595. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-003 Delivered in Joint Session of SEEDS-2015 and COMMUNE-2015

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

5

2015

Keynote – 4 Security Issues in IT Systems and their Management Prof. Abdul Quaiyum Ansari* Department of Electrical Engineering, Jamia Millia Islamia, New Delhi

Keynote IT security means protecting information and information systems from unauthorized access, use, disclosure, disruption, modification, or destruction and IT Security management is a process of defining the security controls in order to protect the information assets. The issue of security related to IT systems is not only technical but is also a governance and organizational issue. Several agencies have been working simultaneously to resolve this issue both at the technical as well as at the organizational levels. Many standards have been developed to handle this complex problem. Every country has its own IT law. The Indian IT ACT 2000 aims to provide the legal infrastructure for e-commerce in India that is supposed to make a major impact for ebusinesses and the new economy in India. This Key Note will focus on the generic issues of the security challenges and their management as also the various perspectives of the IT Act 2000 as related to the international IT security standards and challenges. © 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords: IT ACT; Information Security; IT Law; IT Security Management

*Speaker. Tel.: +91 9873 824597. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-004 Delivered in Joint Session of SEEDS-2015 and COMMUNE-2015

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

7

2015

Keynote - 5 Spin Transfer Torque based Magneto-resistive Memories Dr. Brajesh Kumar Kaushik* Microelectronics and VLSI Group, Department of Electronics and Communication Engineering, I. I. T, Roorkee, Uttarakhand

Keynote Researchers believe spintronics to be one of the most promising technologies to replace the conventional CMOS technology that suffers from severe static leakage beyond 22nm technology node. Spintronics is an emerging technology that exploits an electron's spin orientation and its associated magnetic moment as state variable. It involves the storage of information in terms of non-volatile magnetization state instead of the charge. Thus, the new goal is to develop computing architecture that can normally be off when not in use to prevent static leakage. Moreover, such architecture can be turned on instantly with full performance when required. The primary requisite to achieve non-volatile architecture is non-volatile RAM (NVRAM). Most promising technology to achieve non-volatile RAMs is the emerging spintronics based magneto-resistive memories that switches by spin transfer torque (STT). Spintronics based magneto-resistive memories were revolutionized by the phenomenon of spin transfer torque (STT) effect, first demonstrated by J.C Slonczewski in 1996. After this monumental discovery, spintronics based magneto-resistive memories have evolved considerably in the last decade into their novel form known as spin transfer torque magneto-resistive random access memories (STT MRAMs). STT MRAMs store data as the resistance state of a magneto-resistive device known as magnetic tunnel junctions (MTJs). An STT MRAM cell is composed of two primary components: the "Magnetic Tunnel Junction", which is usually characterized by magneto-resistance and switching current density, and the "Access Device", which allows a given memory cell to be addressed for read or write. This talk will target for a clear understanding of STT MRAMs in terms of architecture, operation and performance comparison with other volatile and non-volatile memory technologies. Moreover, the talk will also focus towards the recent developments and challenges ahead for STT MRAMs. © 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords: Spin Transfer Torque Switching; Magnetic Tunnel Junction; MRAM; Perpendicular Magnetic Anisotropy

*Speaker. Tel.: +91 1332 285662. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-005 Delivered in Joint Session of SEEDS-2015 and COMMUNE-2015

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

9

2015

Keynote - 6 Graphene Based On-chip Interconnects and TSVs: Prospects and Challenges Dr. Brajesh Kumar Kaushik* Microelectronics and VLSI Group, Department of Electronics and Communication Engineering, I. I. T., Roorkee, Uttarakhand

Keynote The conventional on-chip interconnect copper material is unable to meet the requirements of future technology needs, since it demonstrates lower reliability with down scaling of interconnect dimensions. Therefore, researchers are forced to find an alternative solution for interconnects. Graphene nano interconnects have been proposed as promising interconnect materials due to their unique physical properties such as higher thermal conductivity, current carrying capability and mechanical strength. Graphene nano interconnects can be classified into carbon nanotubes (CNT) and graphene nanoribbons (GNR). CNTs are made by rolling up of graphene sheet in a cylindrical form and GNR is a strip of ultra-thin width graphene layer. Most of the physical and electrical properties of GNRs are similar to that of CNTs, however, the major advantage of GNRs over CNTs is that both transistor and interconnect can be fabricated on the same continuous graphene layer. Therefore, one of the manufacturing difficulties in formation of perfect metal-nanotube contact can be avoided. On other hand, the GNRs fabricated till-date, have displayed some level of edge roughness. The electron scattering at rough edges reduces the mean free path (MFP) that substantially lowers the conductance of the GNR. This fundamental challenge limits the performance of GNR interconnects. Presently, researchers and industrialists are standing at crossroads where they need to make subtle improvements to make CNTs and GNRs a workable solution for future. The conventional planar integrated circuit (2D) packaging technique has already hit the red brick wall and is almost on the verge of extinction due to limited number of I/O pins and lower bandwidth. The best way to move towards the “More-thanMoore” technologies is 3D IC packaging, where the dies are vertically stacked. The electrical connections between the dies are established by through silicon vias (TSVs). The idea of using CNTs and GNRs as filler material in TSVs has also rapidly gained research interests. Considering the above-mentioned issues, this talk will analyze and compare the performance of CNTs and GNRs for both on-chip interconnects and TSVs applications. © 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords : Carbon Nanotube; CNT; Graphene Nanoribbon; GNR; On-Chip Interconnects; Through Silicon Vias; TSVs

*Speaker. Tel.: +91 1332 285662 E-mail address: [email protected] Keynote/Expert Lecture ID: SEEDS/COMMUNE-006

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

11

2015

Keynote - 7 A Perso-Arabic to Indic Script Machine Transliteration Model Prof. Gurpreet Singh Lehal* Department of Computer Science, Punjabi University, Patiala, India.

Keynote Indian sub-continent is one of those unique parts of the world where single languages are written in different scripts. This is the case for example with Punjabi, spoken by tens of millions of people, but written in Indian East Punjab (20 million) in Gurmukhi script (a Left to Right script based on Devanagari) and in Pakistani West Punjab (80 million), it is written in Shahmukhi (a Right to Left script based on Perso-Arabic). Whilst in speech, Punjabi spoken in the Eastern and the Western parts is mutually comprehensible in the written form it is not. This is also the case with other languages like Urdu and Hindi (whilst having different names, they are the same language but written, as with Punjabi, in mutually incomprehensible forms). Hindi is written in the Devanagari script from left to right, Urdu is written in a script derived from a Persian modification of Arabic script written from right to left. A similar problem resides with the Sindhi language, which is written in a Perso-Arabic script in Pakistan and both in Perso-Arabic and Devanagari in India. Similar is the case with Kashmiri language too. Konkani is probably the only language in India, which is written in five scripts i.e. Roman, Devanagari, Kannada, Perso-Arabic, and Malayalam. The existence of multiple scripts has created communication barriers, as people can understand the spoken or verbal communication, however when it comes to scripts or written communication, the number diminishes, thus a need for transliteration tools, which can convert text written in one language script to another script arises. A common feature of all these languages is that, one of the script is Perso-Arabic (Urdu, Sindhi, Shahmukhi etc.), while other script is Indic (Devanagari, Gurmukhi, Kannada, Malayalam). Perso-Arabic script is a right to left script, while Indic scripts are left to right scripts and both the scripts are mutually incomprehensible forms. Thus, there is a dire need for development of automatic machine transliteration tools for conversion between Perso-Arabic and Indic scripts. Machine Transliteration is an automatic method to generate characters or words in one alphabetical system for the corresponding characters in another alphabetical system. The transformation of text from one script to another is usually based on phonetic equivalencies. We present Sangam, a Perso-Arabic to Indic script machine-transliteration system, which can convert with high accuracy text written in Perso-Arabic script to one of the Indic script sharing the same language. The system has been successfully tested on Punjabi (Shahmukhi-Gurmukhi), Urdu (Urdu- Devanagari) and Sindhi (Sindhi Perso Arabic - Sindhi Devanagari) languages and can be easily extended for other languages like Kashmiri and Konkani text. © 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords : Perso-Arabic; Machine Transliteration; Indic Script; Indian Languages

*Speaker. Tel.: +91 9815 473767. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-007

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

13

2015

Keynote - 8 Moving Towards “Flawless Telepresence” Systems of the Future Prof. M. Salim Beg* Department of Electronics Engineering, Z. H. College of Engineering & Technology, A.M.U., Aligarh.

Keynote To be able to communicate and compute anywhere and anytime has been one of the goals for last few decades. Due to the huge investments in R&D in the area of Wireless networks in general, and Mobile Communications and Computing in particular, both in the academia and the industry, there has been tremendous outcome in terms of new technologies being developed, as well as resulting products and services. We have in fact now moved on from terms like ‘Ubiquitous Communications’ or ‘Ubiquitous computing’ to newer areas. The traditional mobile and wireless communication networks combined with multimedia communication gave rise to sending text, audio, image, and video to any person, any time, and at any place. In future, several new systems will be made available to us that will lead to something that may be referred to as “flawless telepresence”. The latter implies technologies dealing with not just traditional multimedia elements like text, audio, image, video but incorporating newer elements like smell, touch, and taste. All this will lead to newer frontiers and a paradigm shift in our concept of both computing and communication. Of course, the expectations of users in terms of Quality of Service is growing higher and higher. Thus, one would expect from the service providers to give networks in future that provide richer quality of service and experience by incorporating newer techniques, systems, and networks. While there will be a lot of contributing and enabling technologies for these ‘flawless telepresence’ systems of future, this talk would concentrate mainly on (i) mobile/wireless communication systems and networks (ii) coding and compression of multimedia information and new multimedia systems. Some of the trends and directions in research in these areas will be covered in the talk. © 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords: Flawless Telepresence; Ubiquitous Communications; Ubiquitous computing; multimedia information; Coding

*Speaker. Tel.: +91 9897 023521. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-008

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

15

2015

Expert Lecture - 1 Technology Innovation and Diffusion Practical approach towards 'Make in India' Prof. G. Mohiuddin Bhat* Department of Electronics & Instrumentation Technology, University of Kashmir, Srinagar.

Expert Lecture 'Make in India', an international marketing strategy conceptualized by the Prime Minister of India, Shri Narendra Modi in September, 2014, has been aimed to boost the industrial production and labour intensive manufacturing in India. The objective is to create a job market for a largely unemployed population of 1.2 Billion people in the country, needing about one Million additional jobs every month. The economy of India, presently based on agriculture, cannot be left on the mercy of the unpredictable South-West Monsoons. In order to sustain the rapid growth and alleviate poverty, India rightly needs to harness its potential of 'Make in India'. From the invention of Pencillin to the present day Mobile phone, history is witness to the fact that innovations and inventions have enabled societies to produce more. However, technology innovations can contribute towards productivity only through their application, adoption, and diffusion. Liberal support for technology innovation will enhance entrepreneurship development, which will in turn accelerate the economic growth. Technology innovation and its diffusion are, thus, very crucial towards boosting the manufacturing and service sectors. However, in spite of its large publicly funded science & technology infrastructure, India has not been able to realize its innovative potential. The decrease in the number of indigenous patent applications being filed in India in recent years has raised several questions on the promotion of innovation eco-system in the country. While China topped the global list by filing 5,26,412 Patent applications in the year 2011 (with USA having 5,03,582 patent applications, as runner up), only 42,291 patent applications were filed in India during this period. Realizing that innovation led entrepreneurship development shall promise an economic growth; the Govt. of India has recently taken several initiatives with an innovation agenda. Declaration of 2010-20 as the Decade of Innovation, establishment of National Innovation Council and formulating the Science, Technology & Innovation Policy-2013 (STIP-2013) are indicative of some positive developments in this regard. Further, National level Organizations and programmes like National Innovation Foundations (NIF), Promotion of Innovations among Individuals and MSMEs (PRISM), Grass-root Innovation and Augmentation Network (GIAN) are several other initiatives launched in this direction. Creation of a robust national innovation eco-system is one of the key elements listed in the Science Technology Innovation Policy2013 of the Govt. of India. With a focus on the new initiative of 'Make in India', as conceived by the Prime Minister of India, this article summarizes the possibilities and challenges in the implementation of the initiative through innovation-led entrepreneurship development. The support available for promotion of innovations in the formal and non-formal sectors, as well as the process for protecting innovations through patenting, leading to the diffusion of innovative technologies have been discussed. The recommendations identified in the article shall ensure the flow of technology from academia to the industry, thereby transforming ideas into wealth, and creating a job market in the country.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords: Patent; Innovation; STIP-2013; GIAN; NIF; Make-in-India

*Speaker. Tel.: +91 9906 677322. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-009 Delivered in Joint Session of SEEDS-2015 and COMMUNE-2015

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

17

2015

Expert Lecture - 2 Current State of Cyber Crimes in the State of Jammu and Kashmir

Mr. Rayees Mohammad Bhat* IPS, Superintendent of Police, Hazratbal, Srinagar

Expert Lecture It has been often said that the basic nature of crime is theft... that all crime is theft. Only the dimensions of theft change, the paradigms keep shifting and the convoluted course of time shapes the contours. Every innovation in history has brought a flurry of change in human lifestyle and society. Law and crime being barometers of society have naturally witnessed similar changes. Proverbially crime has stayed one-step ahead of the law. In practice, law has always been a subject of supplydemand dynamics; and once in, it has anticipated a larger slew of likely demands and adapted accordingly. The initiation of technological daily life and the rapid movement along the information highway has led to a generational leap in the past decade. It was pretty rare to find a PC let alone one with an Internet connection then. Now it is a sine qua non. We not only have information literally at our fingertips but we can actually “use” that information in the real sense – buying tickets, comparing prices of that car we like, waiting for online sales to buy our favourite jeans – it’s all happening. And, yet it has brought on its attendant problems of that basic nature of crime – theft. Identity theft, password theft, credit card fraud, bank details phishing, spamming, online cheating, skimming, Nigerian scams, MMS “scandals”, Internet pornography, spy cams, misuse of communication devices, cyber terrorism, cyber wars and what not have become common knowledge nowadays. And, J&K has not been left untouched in this highly evolving scenario. All the above crimes have been perpetrated and are occurring here every day. The trends in cybercrimes show a disconcerting increase over the last few years. Not only cybercrime per se, but usual or traditional crimes unsupported by computer resources and communication devices are now almost unheard of! Emulation of modus operandi and learning criminal / anti-social tricks online has also become a trend. And, J&K Police has been using almost all the tools available at the global level to counter the threats posed by this nouveau aspect in crime.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords: Cybercrimes; Cyber Terrorism; Cyber Wars; Identity Theft; Spamming

*Speaker. Tel.: +91 84918 40107. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-010

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

19

2015

Expert Lecture - 3 Detecting Forgery in Images: A Statistical Perspective Prof. Ajaz Hussain Mir* Department of Electronics and Communication Engineering, National Institute of Technology, Srinagar, India.

Expert Lecture Images have become inseparable part of our life. Almost in all fields ranging from media to evidence in courtrooms, images play an important role. Whereas digitization of images have opened new vistas in image, processing and analysis techniques that enable to extract concealed information from images that may even be beyond visual perception. However, because of the easy availability of photo editing software’s digitized images have, at the same time, become vulnerable to image tampering. Attacker may tamper the images to mislead the public, distort truth, and destroy someone's reputation without leaving any trace. This puts authenticity of any image in doubt. Although techniques like watermarking and stenography can be used to check authenticity of an image but these techniques are no longer viable for every generated image in view of cost in terms of time and complexity. This limitation is overcome by digital image forensics. We need a reliable forensic technique that is able to act as an evidence to image authenticity. A number of forensic image authenticity techniques have been proposed. These work with varying degree of reliability. In our approach, we base our solution on the hypothesis that tampering may change underlying statistics of an image; though traces left by tampering may not be perceptible. It may be pointed out that a number of image forgery techniques exist. However, to test the proposed technique we have used two most commonly used forgery techniques Copy-Move and Splicing on images taken from two standard databases CASIA and CoMoFoD. To test the proposed hypothesis, efficacy of Grey Level Run Length Method (GLRLM) based on second order statistics has been used to detect forgery in images. The features obtained based on GLRLM have demonstrated the potential of proposed method in detecting image forgeries. © 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar.

Keywords: Forgery in Images; Digitized Images; Digital Image Forensics; Watermarking; Stenography

*Speaker. Tel.: +91 9419 010409. E-mail address: [email protected]. Keynote/Expert Lecture ID: SEEDS/COMMUNE-011 Delivered in Joint Session of SEEDS-2015 and COMMUNE-2015

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

21

2015

Articles Title

Authors

Pages

Author Name Disambiguation using a Mix of Hard and Fuzzy Clustering

Tasleem Arif Rashid Ali M. Asger Ghazi Majid Bashir Malik Atul Kumar Kapil Dev Goyal Sajjad Ahmed Mohammad Ahsan Chishti N.A. Kant F.A. Khanday

29-33

M. Tariq Banday Shafiya Afzal Sheikh Javid Ahmad Rather M. Tariq Banday Tawheed Jan Shah Z.A.Bangi F.A.Khanday M. Tariq Banday Farooq Aadil

52-60

Rakesh Prasher Devi Dass Rakesh Vaid Burhan Khurshid Roohie Naaz Mir

78-81

Information Diffusion Modelling and Social Network Parameters (A Survey) Performance Analysis of DPI Overhead on both Elastic and InElastic Network Traffic: A Delay Sensitive Classification and Inspection Algorithm (DSCI)

Mudasir Wani Manzoor Ahmad

87-91

Ashaq Hussain Dar Zubair Manzoor Shah

92-96

Integrated Tactile and Pointing Interface System using NonInvasive Approach

G. Mohiuddin Bhat Rouf Ul Alam Bhat Uferah Maqbool Fayiqa Naqshbandi Naheeda Reshi Fozia Abid Baba

97-102

A Compound of Negative Binomial Distribution with Two Parameter Lindley Distribution as a Tool for Handling over Dispersion

Adil Rashid T. R. Jan Musavir Ahmed

103-109

Grammatical Structure in the Dependency Framework: A Computational Perspective in Kashmiri

Aadil Amin Kak Sumaya Jehangir Mansoor Farooq Sumaira Nabi

110-114

Confusion Matrix based Suggestion Generation for OCR Errors Hybrid Wireless Mesh Protocol in Static IEEE 802.11s Networks Ultra Low-Voltage, Robust and Integrable/Programmable Neural Network based Design of 2:1 Multiplexer File Tracking System for University of Kashmir: Design Guidelines and Model Implementation Color Image Compression using EZW and SPIHT Techniques A Novel Universal (FNZ) Gate Based Adders in QCA Technology A Study of CMOS Frequency Synthesizers in Short Range Wireless Communication A Comparative Study of InSb, InAs and Si based Nanowire MOSFET Optimizing FPGA based Fixed-Point Multiplier using Embedded Primitive and Macro-support

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

34-39 40-44 45-51

61-65 66-70 71-77

82-86

23

2015 Title

Authors

Pages

Phrase Structure in Kashmiri: A UNL Approach

Aadil Amin Kak Sumaira Nabi Mansoor Farooq Sumia Tariq Adil H. Khan T. R. Jan

115-118

Roshani Gupta Rockey Gupta Susheel Sharma Shah Jahan Wani M. A. Peer K. A. Khan Nusrat Parveen Syed Zaffer Iqbal

126-130

Javeed Reshi M. Tariq Banday F. A. Khanday Nadiya Mehraj Faizan Kitab Zia Malik M. Tariq Banday Sukhdev Singh Dharam Veer Sharma Javaid A. Sheikh Shabir A. Parah Uzma Aijaz Tawseef Farah Sanna Aiman G. Mohiuddin Bhat Liyaqat Nazir Roohie Naaz Mir

143-148

Zahid Ashraf Wani Tazeem Zainab

170-174

Muzamil Ahmad Shameem Yousf Sheikh Nasrullah Shifaa Basharat Manzoor A. Chachoo Jayesh C. Prajapti Ekta Khimani Shivani Raval Navdeep Lata Simpel Rani Jindal Deepti Sharma Rakesh Vaid

175-179

Estimation of Stress-Strength Reliability using Finite Mixture of Exponential and Gamma Distributions Design of XOR Gate using Floating-Gate MOSFET

Cellular Automata: Evolution and Parallel Dimensions

High Impedance First-Order Transadmittance-Mode Allpass Filter using CCII and OTA Sequential Circuit Design using Quantum Dot Cellular Automata (QCA) Performance Evaluation of OLSR, DSR and AODV MANET Protocols

Binarization of Natural Science Images Effect of Convolutional Encoding on Bit Error Rate (BER) for Image Transmission using Multiple Input Multiple Output Orthogonal Frequency Division Multiplexing (MIMO-OFDM) System Over Fading Channels.

FPGA Evaluation of Wave Front Allocator for Crossbar based OnChip Switches The Social Web: Expressive use among the Undergraduate Students of University of Kashmir Free Text Plagiarism Detection using Lexical Database and Document Fingerprinting On Linear Classifiers vs. Hybrid Configuration: An Empirical Study Study and Analysis of Downstream ROF PON using TWDM concept Recognition of Typewritten Gurmukhi Characters Effect of Buried Oxide (BOX) in the Drift Region of a Super Junction MOSFET

24

Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar

119-125

131-137

138-142

149-154

155-160 161-164

165-169

180-184 185-188

189-194 195-199

2015 Title

Authors

Pages

On Parameter Estimation of Erlang Distribution using Bayesian Method under different Loss Functions

Kaisar Ahmad S. P. Ahmad A. Ahmed Suriya Jabeen T. R. Jan

200-206

Richa Gupta Deepika Jamwal Rakesh Vaid Harjeet Kaur Rockey Gupta Susheel Sharma Rajandeep Singh M. L. Singh Sajaad Ahmed Lone Faroze Ahmad Renu Rakesh Vaid

214-217

T-X Family of Gamma-Exponential Distribution and Its Structural Properties Impact of Body Thickness on the Performance of InAs Gate-AllAround Nanowire Field Effect Transistor FGMOS based Log Domain Integrator

Optimization of Thulium Doped Fiber Amplifier for S-Band Multimedia Stream Transmission in Mobile Ad hoc Networks

Review on the Electrical Properties of Ultra-Thin Silicon Oxynitride Films Babu Lal Sharma Design of Frequency Reconfigurable Dual Band Microstrip Patch Girish Parmar Antenna with Directional Patterns

Study the Effect of Diode Area on the Current-Voltage and Capacitance-Voltage Characteristics of Al/n-SnSe2/In Thin Film Schottky Diodes Performance Evaluation of Multiplexer Designs in Quantum-Dot Cellular Automata (QCA) Impact of Scaling Gate Oxide Thickness on the performance of Silicon based Triple gate/Quad gate Rectangular-NWFET Detection of Software Cloning by using Visual Detection Technique Support Vector Machine based Multi-Unit Iris Biometric Verification using GLCM for Feature Extraction FloSwitch Board Design using Multi FPGA

On the Realization of Robust Watermarking System for RGB Medical Images

Realization of a Fragile Medical Image Watermarking System for Content Authentication

Content Centric Networking and Interest Flooding in Communication Networks : A Review

Mithilesh Kumar Baljeet Singh Sinwar R.Sachdeva U. Parihar N. Padha

207-213

218-222

223-225 226-231 232-235 236-239

240-244

M. R. Beigh M. Mustafa

245-249

Deepika Jamwal Richa Gupta Rakesh Vaid Harish Patidar Amit Mishra Shiv Kumar Shoaib Amin Banday Ajaz Hussain Mir

250-254

Mursal Ayub Jagannatham V. V Rajshekhar Shabir A. Parah Javaid A. Sheikh Farhana Ahad G. Mohiuddin Bhat Shabir A. Parah Javaid A. Sheikh Zahid Hussain Syed Mohsin Rohit Agnihotri Kshitij Pathak Prashant Bansod Chetan Chouhan

265-269

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

255-259

260-264

270-274

275-279

280-285

25

2015 Title

Authors

Pages

Comparative Performance Analysis of MEMS Piezo Resistive Pressure Sensors Model Order Reduction of Large Scale Continuous Time Interval System

M. Tariq Banday S. Umira R. Qadri

286-291

Rajesh Bhatt Girish Parmar Rajeev Gupta M. R. Dar F. A. Khanday Khalid Sultan Sajad Ahmad Mir Zubida Habib M. Ikram A. H. Moon Ummer Khan Zaffar Kanth Sheikh Junaid Shah Jahan Wani Zahoor Ahmad Peer Fasel Qadir K. A. Khan Shabir A. Parah Javaid A. Sheikh Nazir A. Loan G. Mohiuddin Bhat Javaid A. Sheikh Jai Preet Kour Wazir Shabir A. Parah G. Mohiuddin Bhat M. Y. Kathjoo F. A. Khanday Majid Bashir Malik M. Asger Ghazi Rashid Ali Tasleem Arif Shabir A. Parah Javaid A. Sheikh Jahangir A. Akhoon G. Mohiuddin Bhat M. Tariq Banday Reyaz Ahmad Mathangi

292-298

Javaid A. Sheikh Shabir A. Parah Sakeena Akhtar G. Mohiuddin Bhat I. N. Beigh F. A. Khanday

352-355

Kamal Deep Garg Ajit Kumar Usha Parihar R. Sachdeva C. J. Panchal N. Padha

362-365

Design of a Fractional Order Ramp Generator Structural and Dielectric Studies of RFeO3 (R=Pr, Eu and Ho)

WSN Based Secure Ambient Intelligent Hospitals

QCA Full Adder Design and Noise Problems

A Blind Watermarking Technique in Spatial Domain Using InterBlock Pixel Value Differencing

On the Study and Performance Evaluation of Multirate Filter

Biomedical Sensor Interfacing Circuitry: A Watch Through Privacy Preserving Data Mining using Fuzzy based Approach

High Capacity Data Hiding using Random Plane Indicator Technique for Color Images

Control of IP Address Spoofing - A Comparative Study of IPv4 and IPv6 Networks Performance Evaluation and Comparison of Speech Compression using Linear Predictive Coding and Discrete Wavelet Transform

0.5V Design of Signal Conditioning Circuit for ECG Signal Retrieval A Survey of Spell Checkers Available for Hindi and Punjabi Effect of Semiconductor Thickness on Al/p-CuIn0.81A10.19Se2 Schottky Diodes

26

Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar

299-304 305-308

309-312

313-320

321-326

327-330

331-335 336-338

339-343

344-351

356-361

366-369

2015 Title

Authors

Pages

Question Classification using Knowledge based Semantic Kernel

Mudasir Mohd Zahid Maqbool Shraddha Arya Indu Chhabra G. S. Lehal

370-374

Deepali Vishal Goyal Ajit Kumar Sunita Gaurav Bharadwaj M. M. Sharma Sanjeev Kumar Sharma Gurpreet Singh Lehal Devi Dass Rakesh Prasher Rakesh Vaid Umrinderpal Singh Vishal Goyal Gurpreet Singh Lehal Rohit Sachdeva Dharam Veer Sharma

380-384

Sofi Shabir Roohie Naaz Mir

411-414

G. Mohiuddin Bhat Rouf Ul Alam Bhat Naazira Badar Malik Rabaie Mushtaq Afzan Hussain Hamadani Manzoor A. Chachoo Farah Fayaz Quraishi Summera Ashraf

415-419

Computational Approaches for Emotion Detection and Classification in Textual Data

Abid Hussain Wani Rana Hashmy

425-428

Achievements and Limitation of the First Machine Translation System to Convert Hindi into Dogri

Preeti Devanand

429-432

Computational Aspect of different Neuroimaging Modalities to Study the Structural and Functional Brain Networks: Analysis and Measure Based on Advanced Statistical and Classical Graph Theory Approach

Vinay Shukla Shrawan Kumar Dharmendar Singh

433-435

Study of the Emergence of Sky Computing

Vivek Chalotra

436-440

Digital Identity called Selfie-Means of Narcissism, Self-Exploration or Entertainment? A Review

Aadil Masood Wani Benish Ali Bhat

441-444

Recognition of Offline Handwritten Devanagari Numerals using Statistical Techniques Sentiment Analysis of Views Written in Gurmukhi Script

Rectangular Patch Antenna using Metamaterial for Multi Band Applications Identification of Clause Boundary in Punjabi Language Parameter Optimization of Ballistic Carbon Nanotube Field Effect Transistor Morphological Analysis of Proper Nouns in Punjabi

Issues in Word Segmentation of Handwritten Text in Devanagari Script Remote Monitoring of Water Pollution in Dal Lake using Wireless Sensor Networks in Realtime Machine to Machine(M2M) Control & Communication for Internet of Things (IoT) Using DTMF

Framework for Web Security Using Multimedia Password Authentication

2015 International Conference on Computers, Communication and Electronic Engineering, 16-18 March, 2015

375-379

385-387

388-393 394-397

398-404

405-410

420-424

27

2015 Title

Authors

Pages

Experimental Study of Different Wavelets for Real Time Environmental Monitoring in Wireless Visual Sensor Networks

Umar Farooq Shabir Ahmad Sofi Roohie Naaz Mir

445-450

Search Interfaces of Select Online Public Access Catalogues: An Assessment

Huma Shafiq Sheikh Mohammad Shafi Shazia Rasool Tariq Shafi Zahid Ashraf Wani Samiah Jan Nasti M. Asger Ghazi Muheet Ahmed Butt Majid Zaman Baba Mir Aadil M. Asger Ghazi Vishal Goyal Afshan Amin Khan Shivendra Pandey Joytirmoy Pathak Suresh Kumar Virender Kundu Mamta Nikhil Chauhan Saba Mushtaq Shoaib Amin Banday Ajaz Hussain Mir Anupama Chadha Suresh Kumar

451-455

Padma Prasad Sathisha

486-490

Javaid A. Sheikh Mehboob ul Amin Shabir A. Parah G. Mohiuddin Bhat

491-494

Big Data: A Growing Tide not Hype

English-Kashmiri Machine Translation: Issues and Challenges

A Comparative Analysis of Full Adder Cells in Nano-Scale for Cascaded Applications Synthesis and Characterization of Chemical Bath Deposited CuZnSnS Nano/Microstructures

Verification using Multimodal Biometric Fusion

Extension to the K-Means Algorithm for Automatic Generation of Clusters for Mixed Datasets Implementation of an Embedded Device to Prevent Friendly Fire in Battle Field Improving the Network Capacity and Coverage of 4G-LTEAdvanced Systems using Relaying Technology

28

Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar

456-459

460-464

465-471

472-474

475-479

480-485

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Author Name Disambiguation using a Mix of Hard and Fuzzy Clustering Tasleem Arifa, Rashid Alib, Mohammed Asgerc, Majid Bashir Malikd* a Department of Information Technology, BGSB University Rajouri, Jammu & Kashmir, India Department of Computer Engineering,Aligrah MuslimUniversity,Aligarh, Uttar Pradesh, India c School of Mathematical Science & Engineering, BGSB University Rajouri, Jammu & Kashmir, India d Department of Computer Science, BGSB University Rajouri, Jammu & Kashmir, India b

Abstract Author name ambiguity has long been a problem confronting the effective management of digital literature and digital libraries. Uncertainty about the real authors of a publication sometimes lead to wrong credits to authors or otherwise. Previous studies have tried to solve this problem by using traditional computational techniques. Soft Computing promises to be a good option one can look forward to deal with the problems of uncertainty. In this paper, we present the result of our ongoing work for resolving name ambiguity problem in digital citations. We propose a model that uses both traditional and fuzzy clustering approaches in a two stage framework to solve this problem. The results of our name disambiguation approach which we obtained on DBLP data are very encouraging and we have been able to achieve very good disambiguation performance in comparison to other baseline methods. On an average the values of Precision, Recall and F1 were 94.17, 91.56 and 92.42 percent respectively.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Name Disambiguation; Soft Computing; Fuzzy Clustering; DBLP

1. Introduction The advent of Information Technology has paved the way for proliferation of scientific knowledge and researchers find themselves highly benefited from the use of Information & Communication Technology (ICT) for furthering their research activities (Chang, Huang, 2014). It has been argued (Zhao, Strotmann, 2014) that advances in ICT has led to an increase in research productivity, increased level of research collaborations between researchers geographically far apart from each other, increase in citations, etc. This has also led to accumulation of large amount of bibliographic data in digital libraries like DBLP, CiteSeerX, Microsoft Academic Search, etc. ICT which has made the work of researcher more worthwhile has also compounded the problem for digital libraries by either mixing or splitting the research publications of authors sharing a common name. This is because of the reason that more and more authors with similar names are contributing to scientific knowledge by way of publishing their research work. This is evident from a steep rise in the number of publications in the recent past (Tang, Walsh, 2010). In research publications or bibliographies, the name ambiguity problem arises in two different forms, (a) when same name is expressed in different formats and (b), when different authors express their name in similar ways (Shin et al, 2014). In first case, the ambiguity arises because of not following a uniform naming pattern by an author. This could happen because of different naming conventions by different journals, conferences, book publishers etc. (Han et al, 2003). A case in point is an author Richard Taylor, Professor Emeritus, Information and Computer Sciences, University of California, Irvine. The publications of Richard Taylor appear under six different name variations: Richard N. Taylor; Taylor, R. N.; R.N. Taylor; Richard Taylor; Taylor, R.; and R. Taylor, even on his homepage†, leave aside digital libraries. In second case, the ambiguity arises because of multiple authors sharing a common name (Han et al, 2003). This can happen because of limited number of name options that our parents have while choosing a name for us (Arif et

* Corresponding author. Tel.: +91 9419 182881 E-mail address: [email protected].

ISBN: 978-93-82288-63-3

Tasleem et al/COMMUNE – 2015

al, 2014). In DBLP‡ there are nine different Richard Taylor. Along with the Richard Taylor mentioned in the above example, one Richard Taylor is a senior research fellow at Stockholm Research Institute, one with Institute for Information Policy, College of Communications, Pennsylvania State University, one with University of Houston, etc. These problems have long been impeding the efficient information management and retrieval in digital libraries (Shin et al, 2014). It requires efficient solutions capable of doing correct attribution and classification of publications of ambiguous authors especially when the information available to deal with such a problem is limited and imprecise in some cases. The rest of the paper is organized as follows: in second section, we briefly present the related work; in third section, we present our proposed approach for resolving the name ambiguity problem; in fourth section, we present the experimental results, and in the last section, we conclude the paper. 2. Related Work Efforts for resolving the name ambiguity problem in digital libraries are not new and a number of studies previously have tried to solve the problem broadly using three different techniques: supervised learning (Han et al, 2004; Veloso et al, 2012; Peng et al, 2012), unsupervised learning (Han et al, 2005; Tan et al, 2006; Masada et al, 2007; Soler, 2007; Pereira et al, 2009; Cota et al, 2010) and graphic oriented (Yin et al,2007; Fan et al, 2011). Supervised techniques try to learn a model based on both positive and negative training examples. Han et al. proposed two name disambiguation models, one based on Bayesian probability, and the other on support vector machines. The technique proposed by Veloso et al. uses a supervised rule based classifier. Peng et al. proposed a model based on Web correlations and authorship correlations using a classifier. These methods try to infer the authors of a publication by using various publication attributes like author(s), title, venue, etc. Han et al. (2005) proposed a K-way spectral clustering based name disambiguation mechanism that uses the same kind of information used by Han et al. (2004). The method proposed by Masada et al. uses a two-variable mixture model (by adding two variables), an extension of naïve Bayes mixture model. Another unsupervised model proposed by Soler groups publications iteratively based on the similarity between various publication attributes like author(s), email, title, venue, year of publication, keywords etc. The method proposed by Tan et al. uses a search engine to extract additional information from the Web. On the basis of the information so generated, hierarchical agglomerative clustering (HAC) is used to create clusters of publications. The method proposed by Pereira et al. also obtains additional information from the Web for resolving the author name ambiguity problem. Information is extracted from specific documents on the Web, e.g. CV, by submitting a query to a search engine. The query contains paper title, name of the author and venue. HAC is used to group ambiguous publications which appear on the same Web source. HAC is also used by Cota et al. The clusters are generated in a bottom-up fashion by first fusing them on the basis of similar co-authors, then title of publication and venue of publication. The process is repeated until no more fusions are possible based on the similarity score. The model proposed by Yin et al. applies SVM to weigh different types of linkages used to distinguish authors. In this model what Yin et al. call as DISTINCT, combines two complementary approaches, set resemblance and random walk probability, for measuring similarities between citation records. Another graph theoretic approach, Fan et al. proposed a method called GrapHical framework for name disambiguation (GHOST) using co-authorship information to solve the namesake problem. It first tries to exploit the relationships among publications to construct a graphical model, and solves the namesake problem by serially performing valid path selection, similarity computation, name clustering, and user feedback. GHOST uses only the co-authorship as attribute while excluding all other attributes such as e-mail, publication venue, paper title, and author affiliation, and proposes a novel sophisticated similarity metric to solve the namesake problem. Unsupervised techniques discussed above use hard clustering mechanism, HAC in majority of the cases. None of these approaches make use of fuzzy clustering. To the best of our knowledge no one till date used fuzzy clustering for name disambiguation in digital libraries. The method proposed by us uses a mixture of hard and fuzzy clustering. 3. Proposed Approach Author name disambiguation can be viewed as a classification problem in which it has to be decided whether the publication under consideration belongs to a particular group or not. Classification methods can broadly follow discriminant analysis or cluster analysis technique. Cluster analysis or clustering (commonly known term) is used in those situations where little or no information is available about group structure prior to the classification (Naes, Mavek, 1999). Traditional clustering methods have been used for author name disambiguation in a number of different ways (Arif et al, 2014a). In the proposed approach we use a mix of hard and fuzzy clustering in a two stage clustering framework. In the first stage we use hard clustering framework and in the second we use fuzzy clustering framework. Fig.1 shows ‡ ‡

http://www.ics.uci.edu/~taylor/Publications.htm http://www.informatik.uni-trier.de/~ley/pers/hs?q=richard+taylor

[30]

Tasleem et al/COMMUNE – 2015

the architecture of the proposed system. The bibliographic data for an author name is extracted from DBLP using the methodology shown in fig.2. After extraction this data is supplemented with additional publication attributes obtained in a resource bound manner (Kanani et al, 2007) from WWW using a search engine. We do not go into the details of the extraction of the additional publication features. In first stage we use the clustering process and similarity measures used in (Arif et al, 2014). In second stage, we compute the distance between all the available attributes of a publication with those of the other to combine these distances into a similarity score. This similarity score is used to calculate the value of membership function used for fuzzy clustering step.

Fig.1: Architecture of the proposed system.

Fig.2: Publications Data Extraction from DBLP

Fuzzy or soft clustering allows data elements to belong to more than one cluster simultaneously, and be associated with each cluster with certain membership levels. The degree or grade of membership which can be any value in the range [0, 1] indicates the strength of the association between that data element and a particular cluster. Soft clustering is a process of assigning these membership values, and then using these membership values to assign data elements to one or more clusters. Soft clustering has proved to be beneficial in dealing with uncertainty. There may be certain cases where agglomerative clustering used in first stage may have a good number of clusters with only one citation record. In such a case we use fuzzy clustering to find the relative similarity between clusters having a single publication and the rest by calculating the value of membership function (µ) by using equation (1) as follows:

cos cri , c j 

m

 i, j 



coscri , c r  r 1 k

m

(1)

where cri and Cj are ith publication in a single publication cluster and jth cluster (i ≠ j), respectively, and m is the fuzzy factor. The parameter m determines the “softness” of the clustering solution. If m=0, the degree of membership of a publication with all the remaining clusters is same and when m approaches ∞, the clustering becomes hard clustering (Zhao, Karypis, 2004). In general, the softness of the clustering solution is inversely proportional to fuzzy factor m. In our case, we merge a singleton cluster with any other cluster only if the value of fuzzy membership function is above a threshold. 4. Experimental Results We extracted bibliographic data from publications of ten ambiguous authors indexed by DBLP. The statistics of the dataset used are shown in Table-1. Here, #Publications refer to the number of publication records retrieved from DBLP for the author name listed in a particular record, #Actual Authors to the number of real authors, #Predicted Authors-HC to the number of authors predicted by the proposed approach after hard clustering stage and #Predicted Authors-FC to the number of authors predicted after fuzzy clustering stage.

[31]

Tasleem et al/COMMUNE – 2015 Table.1: Dataset Statistics. Author Name Jim Smith Ajay Gupta Michael Wagner Rakesh Kumar Hui Fang Jie Tang Richard Taylor Paul Jones Robert Fisher Gang Wu

#Publications

#Actual Authors

81 107 232 233 156 227 186 57 191 295

6 11 20 33 21 12 19 11 11 24

#Predicted AuthorsHC 6 14 25 31 25 13 24 13 17 40

#Predicted AuthorsFC 6 13 19 28 23 12 20 13 16 37

The performance of the proposed disambiguation approach used in this study has been shown in terms of percentage Precision, Recall and F1 scores in table-2. Table-2 presents the values of the above metrics for all the authors and the percentage change in the values of these metrics over their respective values obtained in the first stage i.e. hard clustering stage. Although these values may not represent any significant change but whatever they have been able to achieve is quite meaningful for the disambiguation process. In case of Michael Wagner, the fuzzy clustering step has been able to improve the values of precision, recall and F1 by a margin of 3.415, 5.278 and 4.368 percent respectively. On an average the improvements in precision, recall and F1 for all the ten authors is 0.86, 0.00 and 0.40 percent. We considered HAC (Tan et al., 2006), which uses agglomerative clustering and the metadata information is augmented using search engine results in a similar fashion that we used in the first stage, for comparison of disambiguation results. The comparison of the results obtained through the proposed approach with the base line method taken from (Tang et al., 2012) on all three metrics listed above is shown in table-3. The values of precision, recall and F1 obtained through the proposed approach were 94.17, 91.56 and 92.42 percent, respectively. In case of majority of the authors under consideration, the values of all the three metric were more than 90 percent. In case of Gang Wu, the low value of recall was instrumental in bringing down the value of F1 to 67.28. The low value of recall in this case can be attributed to large number of false-negative cases as more than one Gang Wu published in a similar venue. The proposed approach has been able to improve the values of precision and F1 by 21.55 and 13.78 percent respectively over HAC. However, the value of recall decreased by 0.13 percent. Table.2: Values of Precision, Recall and F1 and their percentage change in comparison with results of first stage i.e. hard clustering. Final Results (After Fuzzy Clustering) Precision Recall F1 Jim Smith Ajay Gupta Michael Wagner Rakesh Kumar Hui Fang Jie Tang Richard Taylor Paul Jones Robert Fisher Gang Wu Average

100.00 97.12 100.00 84.51 98.04 98.23 90.45 88.46 95.81 89.09 94.17

100.00 96.19 95.50 89.11 97.40 99.55 93.60 90.20 100.00 54.04 91.56

Change over Hard Clustering (%) Precision Recall F1

100.00 96.65 97.70 86.75 97.72 98.89 92.00 89.32 97.86 67.28 92.42

0.000 0.029 3.415 1.031 0.671 0.442 0.625 0.000 0.549 1.818 0.86

0.000 0.038 5.278 -1.433 -0.636 -0.448 -1.715 0.000 0.000 -1.103 0.00

0.000 0.033 4.368 -0.168 0.015 0.000 -0.525 0.000 0.281 0.000 0.40

Table.3: Comparative F1 Scores (%age) of Representative Authors with HAC

Jim Smith Ajay Gupta Michael Wagner Rakesh Kumar Hui Fang Jie Tang Richard Taylor Paul Jones Robert Fisher Gang Wu Average

Prec.

HAC Recall

F1

Prec.

Our Approach-MSC Recall

F1

92.43 41.88 18.35 63.36 100.00 100.00 80.17 36.36 96.14 97.54 72.62

86.80 100.00 60.26 92.41 100.00 100.00 99.93 80.00 100.00 97.54 91.69

89.53 59.04 28.13 75.18 100.00 100.00 88.97 50.00 98.03 97.54 78.64

100.00 97.12 100.00 84.51 98.04 98.23 90.45 88.46 95.81 89.09 94.17

100.00 96.19 95.50 89.11 97.40 99.55 93.60 90.20 100.00 54.04 91.56

100.00 96.65 97.70 86.75 97.72 98.89 92.00 89.32 97.86 67.28 92.42

[32]

Tasleem et al/COMMUNE – 2015

5. Conclusions Resolving name ambiguity has become important in view of increasing number of publications and widespread usage of digital libraries among the researchers. In this paper we proposed a hybrid clustering mechanism by employing hard clustering in first stage and soft clustering in the second. Experimental results conducted on DBLP dataset are very encouraging as the proposed approach has been able to achieve F1 score of 92.42 percent. By using soft clustering we have been able to deal with the split citation problem to a good extent. In some cases where F1 score were below expectations is due to the fact that more than one authors published with same journals or conferences which lead to distinct clusters being merged based on venue information. We are also of the view that with ever-increasing number of publications with similar authors venue information may not prove to be a good feature for disambiguation purposes. References Arif, T., Asger, M., Ali, R. (2014a). Author name disambiguation using two stage clustering. INROADS-An International Journal of Jaipur National University (Special Issue), ISSN: 2277-4904, Vol.3,no.1, pp-340-345. Arif, T., Ali, R., Asger, M. (2014b). Author name disambiguation using vector space model and hybrid similarity measures. In Proceedings of 7th International Conference on Contemporary Computing-IC3’2014, Noida, India: IEEE. pp: 135-140. Cota, R.G., Ferreira, A.A., Nascimento, C., Gonçalves, M.A., Laender, A.H.F. (2010). An unsupervised heuristic-based hierarchical method for name disambiguation in bibliographic citations. Journal of the American Society for Information Science and Technology, Vol.61, no.9, pp: 1853– 1870. Chang, H.W., Huang, M.H. (2014). Cohesive subgroups in the international collaboration network in astronomy and astrophysics. Scientometrics, Vol.101, no.3, pp: 1587-1607. Fan, X., Wang, J., Pu, X. Zhou, L., LV, B. (2011). On graph-based name disambiguation. ACM Journal of Data and Engineering Quality, Vol.2, no.2, pp: 10. Han, H., Zha, H., Giles, C.L. (2003). A model-based K-means algorithm for name disambiguation. Proceedings of 2nd International Semantic Web Conference, USA. Han, H., Giles, L., Zha, H., Li, C., Tsioutsiouliklis, K. (2004). Two supervised learning approaches for name disambiguation in author citations.” In Proceedings of Joint Conference on Digital Libraries’2004, pp: 296 – 305. Han, H., Zha, H., Giles, C.L. (2005) Name disambiguation in author citations using a K-way spectral clustering method. In Proceedings of Joint Conference on Digital Libraries’2005, pp: 334 – 343. Kanani, P., McCallum, A., Pal, C. (2007). Improving author coreference by resource-bounded information gathering from the web. Proceedings of 20th International Joint Conference on Artificial Intelligence-IJCAI, Hyderabad, India, pp: 429-434. Masada, T., Takasu, A., Adachi, J. (2007). Citation data clustering for author name disambiguation. In Proceedings of 2nd International Conference on Scalable Information Systems. Naes, T., Mevik, B-H. (1999). The flexibility of clusters illustrated by examples. Journal of Chemometrics, 13(-4), pp: 435-444. Peng, H., Lu, C., Hsu, W., Ho, J. (2012). Disambiguating authors in citations on the web and authorship correlations. Expert Systems with Applications, Vol.39, no.12, 10521-10532. Pereira, D.A., Ribeiro-Neto, B., Ziviani, N., Laender, A.H., Gonçalves, M.A., Ferreira, A.A. (2009). Using web information for author name disambiguation. In Proceedings of 9th ACM/IEEE-CS Joint Conference on Digital Libraries’2009, ACM. Shin, D., Kim, T. Choi, J., Kim, J. (2014). Author name disambiguation using a graph model with node splitting and merging based on bibliographic information. Scientometrics, Vol.100, no.1, pp: 15-50. Soler, J. (2007). Separating the articles of authors with the same name. Scientometrics, Vol.72, no.2, pp: 281-290. Tan, Y.F., Kan, M. and Lee, D. (2006). Search engine driven author disambiguation. Proceedings of ACM/IEEE Joint Conference on Digital Libraries (JCDL ’06), pp: 314-315. Tang, L., Walsh, J.P. (2010). Bibliometric fingerprints: name disambiguation based on approximate structure and equivalence of cognitive maps. Scientometrics, Vol.84, no.3, pp: 763-784. Tang, J., Fong, A.C.M., Wang, B., Zhang, J. (2012). A unified probabilistic framework for name disambiguation in digital library. IEEE Transactions on Knowledge and Data Engineering, Vol.24, no.6, pp: 975-987. Veloso, A., Ferreira, A. A., Gonçalves, M. A., Laender, H.F.A., Meira Jr., W. (2012). Cost-effective on-demand Associative Author Name Disambiguation. Information Processing and Management,Vol. 48, no.4, 2012, pp: 680– 697. Yin, X., Han, J., Yu, P.S. (2007). Object distinction: Distinguishing objects with identical names. In Proceedings of IEEE International Conference on Data Engineering, pp: 1242-1246. Zhao, D., Strotmann, A. (2014). The knowledge base and research front of information science 2006–2010: An author cocitation and bibliographic coupling analysis. Journal of the Association for Information Science and Technology, Vol.65, no.5, pp: 995–1006. Zhao, Y., Karypis, G. (2004). Soft clustering criterion functions for partitional document clustering: a summary of results. Proceedings of 13th ACM International Conference on Information and Knowledge Management, New York, USA, pp: 246-247.

[33]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Confusion Matrix Based Suggestion Generation for OCR Errors Atul Kumar, Kapil Dev Goyal* Department of Computer Science, Punjabi University, Patiala, India

Abstract This paper proposes a method for generating suggestion of errors done by OCR system based on confusion matrix for Punjabi language (Gurumukhi script). Confusion matrix is developed from large text Corpus. The proposed method firstly determines the probability of confusion of one character (OCR output) with another character (OCR input) from confusion matrix. For each word of OCR output, number of strings is generated from generated lists of topmost five confused characters for each character of input word and probability scoring of these strings is calculated for ranking. While generating strings, each string is validated with the trigram lexicon. If validated, that string is taken and after generating all the valid strings, lexicon is used for best suggestions. The topmost five words are taken as suggestions. This method has been tested for variety of OCR outputs documents. The system developed also used for Devanagri Script.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: DWT; DCT; Embedded Zero Tree; Set Partitioning in Hierarchical Tree

1. Introduction With the improvement in technology, current OCR technology is rectified but there are many old documents like books, text documents, which are of poor quality results in many errors during OCR processing. The post-processing is the last stage in OCR system. The main purpose of post-processing is to remove errors occurs during the various stages of OCR. Mainly the errors occur between the characters, which have some similar shape structures (like in Gurumukhi script ਥ with ਖ). Correction of these kinds of errors is very difficult one. The main aim of this research work is to develop a system that perform the post-processing based on confusion matrix and generates the spelling suggestions without taking into account the context. The rest paper has been organized as follows: Part 2 discusses previous work in the field of post-processing of OCR; Part 3 explains the proposed research work, Part 4 shows experimental results, Part 5 gives conclusion about research work and Part 6 contains References. 2. Previous Work Post-processors have been extensively used in the past to enhance OCR accuracies. Various OCR post-processing techniques are used from past decades. (Kernighan, et al, 1990) developed a program for spelling corrections based on a Noisy channel which proposes a list of candidate corrections and sort them according to their probability (Kenneth, et al, 1991) developed the new program called spell which took the incorrect word and proposed a list of candidates words along with probabilistic scoring. The performance of system was increased up to 89%. (Kukich, 1992) has done the survey on various kinds of errors and explained various kinds of techniques used in this area. Masaaki NAGATA used the statistical OCR model, an approximate word matching method using character shaped similarity and word segmentation algorithms. The accuracy of 90% to 97.34% has been found. (Kolak, et al, 2005) proposed lexicon free method for error correction and implemented using FSM (finite state machines). Authors have claimed the improvement of about 78% error reduction obtained for low density languages like Igbo. (Bassil, et al, 2012) used Google’s spelling suggestion scheme, which is based on the probabilistic n-gram model for predicting the next word in * Corresponding author. Tel.: +919814877883. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Kumar and Goyal/COMMUNE – 2015

a particular sequence of words. The error rate was dropped up to 3-4% by applying these algorithms. In case of Indian Language script for OCR systems, (Pal, et al, 2000) proposed a technique which was based on morphological parsing using two separate lexicons of root words and suffixes. Authors have claimed the accuracy of 84.27%. Lehal, et al, 2001) developed a shape based postprocessor for Gurumukhi OCR. The recognition accuracy of the OCR without post processing was 94.35%, which was increased to 97.34% on applying the post-processor to the recognized. Sharma et al developed shape encoding based post-processing for Gurumukhi (an Indian script).The accuracy of 4-7% was seen. (Mohan, et al, 2010) proposed a post-processing scheme which use statistical language models at the sub-character level to boost word level recognition results. Authors have claimed the accuracy rate of 92.67% for Malayalam text. (Jain, et al, 2011) proposed used the minimum edit distance technique along with substitution cost as confusion probabilities of characters. Authors have claimed to improve the accuracy of OCR by 33%. 3. Proposed Solution As mentioned in second part, various methods are there to correct the OCR errors. This paper proposes a method to generate suggestions of OCR errors based on confusion matrix. The proposed method firstly creates a confusion matrix from a large corpus text. 3.1.

Confusion matrix

Confusion matrix shows how many times one character is confused with other character. The accuracy is obtained by dividing the no. of correctly recognized characters to the total number of characters images which are actually present in the database. The confusion between character recognition is due to shape similarity of the characters. Handwritten data increases the confusion further. In order to get the accuracy, confusion matrix is obtained from large corpus of data. Table 1 shows the confusion matrix for Gurumuhi script. The first row shows actual characters in OCR input and first column shows the confused characters with which actual characters are confused. Each cell represents how many times particular row character confused with column character. Tacle 1. Confusion Matrix of Gurumukhi Script.

3.2.













136828

32

11

0

2



0

37795

32

0

0



70

231

87324

1

0



2

5

0

44012

2



16

0

3

4

26481

Calculation of probabilities from confusion matrix

This step involves the calculation of probabilities i.e. what is the probability of one character confused with other character. The probabilities are calculated from the confusion matrix by using the following formula

pr(y/x ) = num(sub(x,y))/num(x) Where pr is probability, num (sub(x,y)) is number of times x is substituted by y and num(x) is the total number of x. For example suppose ਅ is replaced by ਆ. So from table 2. pr(ਆ / ਅ ) = num(sub(ਅ/ ਆ)/num(ਅ) = 231/38575 (calculated) =0.00598(Result taken from implementation) Table 2. .Confusion matrix showing probabilities of confusion of column character with each row character.













0.98830

0.0008

.0001

0

0.000075



0

0.9798

.0003

0

0

[35]

Kumar and Goyal/COMMUNE – 2015

3.3.

Creating Top five confusion Probabilities

Since in confusion matrix, there are many entries, which have very low probabilities as shown in table.2, so we have to ignore those probabilities along with characters. Table 3. Confusion Probability of ਖ

3.4.

Characters ਖ

Confusion Probability 0.846768456192196



0.0862201061675696



0.0452033054233021



.00448749521151426



0.00311935642751601

Creation of Character Trigram Lexicon

For validating word, generally the method is to look in the dictionary. But the problem is that there are not all the words in the dictionaries. In addition, no forms of verbs are available in dictionaries. For e.g. ਦੀ, ਦਾ, ਦੈ. All these words are not available but if we generate character trigrams then all these words are validated. Steps for creating trigram dictionary are: (i) For each unique word in the dictionary we create character trigrams, e.g.: ਸ਼ਹਿਰ ਸ਼ਹਿਰ

ਰ@@

@@ਸ਼ @ਸ਼ਿ ਸ਼ਹਿ

ਿਰ@

ਹਿਰ

Fig. 1.Character trigrams for ਸ਼ਹਿਰ

We have taken the dictionary of 105629 words. (ii) Each trigram generated above is stored in Binary Search Tree to avoid duplicity. (a) Generate trigram of the word as mentioned earlier. (b) Search for the trigram in the BST. (c) If trigram is found in the BST tree then reject it (d) If trigram is not found add a new node with that trigram. Finally, In-order traversal is applied on the BST tree and traversed nodes are stored in a file. We have applied this method for around 105629 words and 23031 character trigrams are obtained. 3.5.

Generation of words and Creation of Suggestions

When post processing is to be applied, first the character trigram Lexicon (Fig 1.) is loaded into Binary search tree. Also simple dictionary is loaded in binary search tree. The following algorithm is applied for generating suggestions for each word of OCR output: (i) Let S be the source word which is taken from OCR output. (ii) Firstly calculate the length of word. After that dynamically generate number of lists equal to the length of word. (iii) For each character of word S, generate the dynamic array list containing the top five confusing characters along with probabilities. Characters are stored at even positions starting from zero and corresponding probabilities at odd positions.

[36]

Kumar and Goyal/COMMUNE – 2015

(iv) After the lists are generated for each character of source word S, next step is to generate words by combining each character of every list with every character of other lists of source word S and also generating probabilities for the purpose of ranking. 3.5. 1.

Validation Process

Before generating suggestions, we have to validate each word. For each word generated in step (iv)  Generate the character trigrams as shown in figure 1.  Search each trigram in trigram dictionary loaded as Binary search Tree (Fig 2). If all the trigrams of word are present in Binary search tree then word is valid otherwise invalid. After this we used dictionary to remove wrong words. Each word generates above is searched in the lexicon. Dictionary we have taken contains around 105629 words. If the search is successful then that word is placed in final suggestion otherwise deleted. OCR text that is to be corrected

Confusion Matrix

Candidate generation with probability

Lexicon

Candidate validation

No

Character Trigram Lexicon

Validated

Yes Candidate Tree Generation

Top five Suggested Candidates

Corrected Text Fig.2. System Architecture

(v) The valid words are ranked according to probabilities and top five probabilistic valid words are taken as suggestions. For the efficiency of the Post-processing, it is important that the right suggestion is presented at the top position. Otherwise, the efficiency of system will affect. So the suggested words are again checked against Lexicon to remove non words and to improve the positions of suggestions so that correct word can be found at rank 1 or rank 2. 4. Experimental Results We have tested the performance of this post-processing on variety of OCR output documents. Following are the initials taken  Number of distinct words in dictionary = 105629  No. of nodes created for dictionary in Binary search tree=105629  No of character trigrams generated from the dictionary words = 23030

[37]

Kumar and Goyal/COMMUNE – 2015

For correct words we have taken around 3400 words from 40 OCR documents and accuracy is measured 100%. Fig 3 shows results. Here First position shows suggestion at top with probability. Second position represents suggestion at second position and so on. Not found means correct word is not obtained. The vertical bar represents number of words related to following category. Example ਹਿਆ is correct word correctly at tops along with highest probability as shown in table 4. Table 4. ਹਿਆ suggestions along with probabilities

Rank.

Word

Probability

1

ਹਿਆ

0.984526219030382

2

ਹਿਆ

0.000799556482201164

3

ਹਰਆ

0.000535872961475248(

4

ਹਿਆ

0.000484837441334748

5

ਹਿਆ

0.000365754561006916.

Correct words suggestions 4000 3000 2000 1000 0 Correct words

Fig.3. Correct words suggestions for correct words

For incorrect word are taken from same 40 OCR documents and firstly find performance of developed system without applying dictionary. Fig 4 shows results. The donations of bar chart are same as mentioned in previous chart. Example ਪਰੀਹਥਆਢਾਂ is incorrect word. The correct word is at position 4 shown in Table 5 Table 5. ਪਰੀਹਥਆਢਾਂ suggestions along with probabilities

Rank.

Word

Probability

1

ਪਰੀਹਬਆਵਾਂ

0.00105692891130884

2

ਪਰੀਹਬਆਦਾਂ

0.00054927014288491

3

ਪਰੀਹਬਆਚਾਂ

0.00035785782036441

4

ਪਰੀਹਖਆਵਾਂ

0.000238979264328572

5

ਪਰੀਹਖਆਦਾਂ

0.000124193948391227.

[38]

Kumar and Goyal/COMMUNE – 2015

Fig.4. Correct words suggestions for incorrect words

When dictionary is applied, then we have seen the improvement of first position suggestions that are correct words. Also 4th and 5th position suggestions are diminished. The overall accuracy has been improved. Table 5 shows the results. Example ਪਰੀਹਥਆਢਾਂ is incorrect word. When lexicon is not applied, correct word is found at 4 th position as mentioned in Table 4.When dictionary(lexicon) is applied , correct word is at top position as shown in Table 4

Fig.5 Correct words suggestions for incorrect words using Lexicon

5. Conclusion The developed system uses different sources like character trigrams, Lexicon dictionary, and confusion matrix. It corrects non words as well as words which have some meanings. The accuracy rate of OCR results has been improved up to 96.14%.This system does not take care of name inconsistency which can be a subject of further research. References Bassil, Youssef, Alwani. Mohammad, 2012. OCR post-processing error correction algorithm using Google’s online spelling suggestion, Journal of Emerging Trends in Computing and Information Sciences, p, 90-99. Jain. Rupi, Chaudhury.Santanu.2011.Probabilistic Approach For Correction Of Optically-Character-Recognized Strings Using Suffix Tree in Proceedings of Third National Conference on Computer Vision, Pattern Recognition, Image Processing and Graphics ,p, 74-77. Kenneth. W. Church, William. A. Gale, 1991. Probability scoring for spelling correction, Statistics and Computing, Volume 1, Issue 2, p, 93-100. Kernighan. D. Mark, Kenneth. Church. W, Gale. A. William, 1990. A Spelling Correction Program Based on a Noisy Channel Model. AT&T Bell Laboratories 600 Mountain Ave Murray Hill, N.J., USA, p. 205-210. Kolak. Olan, Resnik. Philip, 2005. OCR Post-Processing for Low Density Languages. Computer Science and UMIACS, University of Maryland College Park, MD 20742, p. 867-874. Kukich. K, 1992.Techniques for Automatically Correcting Words in Text in ACM Computing Surveys , Vol. 24, No. 4, p, 377-439. Lehal. S. G, Singh. Chandan, Lehal.Ritu.2001. A Shape Based Post Processor for Gurumukhi OCR, Proceedings of the Sixth International Conference on Document Analysis and Recognition (ICDAR’01) IEEE Computer Society Press, USA , p.1105-1109. Mohan. Karthika, Jawahar. V. C., 2010. A Post-Processing Scheme for Malayalam using Statistical Sub-character Language Models in Ninth IAPR International Workshop On Document Analysis Systems, Boston, MA ,p. 493-500. Nagata.. Masaaki, 1998. Japanese OCR Error Correction using Character Shape Similarity and Statistical Language Model. in Proceedings of the 36th annual meeting on Association for Computational Linguistics -Volume 2 , COLING-ACL, p.922 – 928. Pal.U, Kundu. K. P, Chaudhuri. B.B., 2000. OCR Error Correction of an Inflectional Indian Language Using Morphological Parsing. Journal of Information Science and engineering, p, 903-922 Sharma. Veer. Dharam, Lehal. G. S, Mehta.Sarita.2009. Shape Encoded Post Processing of Gurumukhi OCR, Proceedings of tenth International Conference on Document Analysis and Recognition p. 788-792.

[39]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Hybrid Wireless Mesh Protocol in static IEEE 802.11s Networks Sajjad Ahmeda, Mohammad Ahsan Chishtib* a

Department of Information Technology, National Institute of Technology Srinagar, India Department of Computer Science Engineering, National Institute of Technology, Srinagar

b

Abstract Wireless mesh network (WMN) is an important technology which is being used in next generation wireless networks to solve last mile problem of anywhere, anytime and low cost connectivity . IEEE has developed an extension of IEEE 802.11 Wireless network based on wireless mesh networks. The standardization work is in progress and is called IEEE 802.11s networks. The default routing protocol specified in the draft is Hybrid Wireless Mesh Protocol (HWMP) and airtime metrics as default routing metric. In this paper, author analyzed the performance of modes of Hybrid Wireless Mesh Protocol (HWMP) using Network Simulator -3(NS3).

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Routing protocol; wireless Mesh network; IEEE 802.11s; hybrid Wireless Mesh Protocol; Network Simulator

1. Introduction Wireless local area networks (WLANs) are one of the most popular technologies for providing network services are as compared to their wired counterparts because of ease, low cost deployment and their support for user mobility. WLANs are growing rapidly and are found everywhere. One of the reasons for their widespread use is the unlicensed ISM band over which it operates. However, ISM band is interference limited. Therefore, maximum allowable output power is regulated as described in IEEE 802.11 Standard which limits range of IEEE 802.11 based devices. To extend coverage further single hop communication is replaced with multi-hop communication. The Multihop communication can help extend the transmission coverage area without installing new access points as described in (Seungjoon Lee et al, 2004). To add Multihop capability to existing standard IEEE Task Group “S” is working on a new standard called IEEE 802.11s, which is in draft stage. Few implementations are still available such as OLAP, openMesh and open80211s by cozybits. The amendment being considered is based on Wireless Mesh Networks (WMN). Detailed Survey on WMN can be found in the paper by (Akyildiz at el, 2005). The Main hurdle for wide spread use of wireless communication networks was low data rate provided by wireless communication system as compared to wired networks but with recent advances in wireless communication technology wireless networks are providing high data rates as described in (Kuran et al, 2007). As a result, wireless Internet access is everywhere these days. Network Deployment is carried out by placing wireless access points at well-planned places. As per IEEE 802.11s Wireless LAN Mesh Networking standardization, WMNs additionally simplifies network establishment, administration, and maintenance. It also makes setup of networks more cost effective as mentioned in Raffaele Bruno et al. The properties such as cost effective, easy to deploy, establish, and maintain make WMN based networks ideal for emergency scenarios. As compared to traditional WLANs WMN based networks requires and planning as well as less administration as described in (Camp et al, 2008). Traditionally routing protocols are implemented at network layer to enable multihop forwarding, which makes the approach link layer independent. A recent proposal in wireless networking puts multihop forwarding in the link level, which extends the WLAN functionality of nodes. For better utilization of wireless links additional quality aware metrics which have access to link level parameters of the channel can be implemented at MAC level. Currently, The IEEE Task Group 802.11s is developing a promising standard for mesh networking at the MAC level. This new approach provides a large number of benefits. Such an approach makes it appear as a LAN for layerthree protocols. Implementing multihop routing at network layer need modify TCP\IP layer which is not required if multihop is implemented at link layer. IEEE 802.11s specifies a mandatory routing protocol Hybrid Wireless Mesh *Corresponding author. Tel.: +91 9419 023039. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Ahmed and Chishti/COMMUNE-2015

Protocol (HWMP), and a mandatory route metric for path selection called Airtime Metric. The standard also specifies a framework for implementing new path selection and path metric for future extensions, Guido R Hiertz et al. In IEEE 802.11s networks, routing is implemented at layer 2. Thus, high-level layer are shielded from the wireless mesh network at MAC level. The routing at MAC level is called Path selection (Carrano et al, 2011). The paper is organized as follow: section 2 contains background about the routing protocol hybrid wireless mesh protocol (HWMP), section 3 contains simulation setup details, results are presented in section 4 and conclusion is presented in section 5. 2. Hybrid Wireless Mesh Protocol Hybrid Wireless Mesh Protocol is a routing protocol implemented at layer-2 that is MAC Layer. The protocol is a hybrid of a reactive and a proactive routing protocol. The reactive part of the protocol is a variant of Adhoc On-demand Distances Vector (AODV) Routing protocol called Radio Aware-AODV. The proactive part of the protocol is a tree based protocol. The Combination of two approaches helps in finding an optimal and efficient paths in a wide variety of networking scenarios. Routing or path selection table is stored at MAC layer instead of IP layer. Thus, hiding the Multihop nature of MAC layer from upper layers as described in Chun-Ping Wang et al. HWMP uses airtime metric for link measurement. To differentiate between routing at network layer and routing at layer 2 the term path selection is used instead of routing. A mesh node which participate in path selection is known as Mesh Point (MP). The control messages used by HWMP are Path Request (PREQ), for finding new paths, Path Reply (PREP), packets sent by destination or by those having path to destination, in response to PREQ, Path Error (PERR), packet sent to source when paths are no more available. Route Announcement (RANN) packets are flooded into the network for announcing routes. A special metric field is used to propagate path metric information between Mesh Points(MP).Sequence number are maintained by MPs to avoid infinite loops and are exchanged with other MPs through HWMP control messages. 2.1.

Modes of HWMP

HWMP supports two modes of operations, Reactive Mode and Proactive Mode. The Proactive mode is further divided into three types viz. Simple Proactive Mode, Forced Proactive Mode, Proactive With RANN. In reactive mode a path to destination node is determined as and when needed. No tables are maintained to store paths. To find a route or path, source MP broadcasts a special routing control message called Path Request message (PREQ) with the destination field of the frame is initialized with the destination address and the metric field initialized to 0. The destination MP on receiving PREQ replies with path reply message (PREP). The intermediate nodes may be allowed to reply to the source node depending on whether a particular flag called “destination only” flag is enabled or not. In proactive mode, path to every mode is determined beforehand so that a latency free transmission of data can be commenced immediately. In this mode, a special root is designated as root mesh station which acts as Gateway for external traffic. A root mesh station proactively creates a tree with root mesh station as root of the tree and other mesh station end points of the tree. As a result a path to every mesh station is always available. Periodically root broadcasts Path request message (PREQ), telling every mesh station about the root. There are two proactive modes of building path trees in HWMP, Simple and forced proactive mode. In simple proactive mode, the mesh stations are not allowed to send Path reply message, PPREP, in response to the path request, PREQ, broadcasted by root mesh station. Every mesh station learns about the path to root and update their metrics toward root. As a result a unidirectional paths are created from all nodes to the root node. Suppressing of PREP messages is achieved by setting the Proactive Path Reply flag of the proactive PREQ control message. No proactive paths are available from root node to other nodes in the network. In case root node needs a path to some mesh node reactive approach is used to select path. In forced proactive mode bidirectional paths exists between root and every mesh station. This is achieved by setting path reply flag in the PREQ message forcing every mesh station to reply to the PREPs sent by the root. In effect a bi directional tree is created between root and every mesh station enabling both root and mesh stations to start latency free transmission. In IEEE 80.211s airtime metric is identified as default path metric. Airtime metric is given preference over other metrics as it is an efficient radio-aware metric as described in Michael Bahr et al. It represents the amount of channel resources consumed for transmitting a frame over a particular link. Best path is the path over which a frame occupies transmission medium for less duration as transmission medium in wireless communication is expensive resource. 3. Simulation The Network Simulator-3 (NS-3) was used to study the performance of modes of HWMP. Network Simulator is a discrete-event network simulator with a special focus on Internet based systems. The ns-3-dev version was used which is a development version containing the latest release of this network simulator. Wireshark, patched with mesh patch,

[41]

Ahmed and Chishti/COMMUNE-2015

was used for analysis of captures packet data In NS-3 only two modes, reactive and forced proactive, of HWMP are implemented. The simple proactive mode was implemented by editing code in hwmp-protocol.cc. After doing the changes and compiling the ns-3-dev, the simulation work was started by implementing evaluation scenario. 3.1.

Evaluation Scenario

Scenario used for the evaluation of performance of HWMP models a typical WLAN installed in a campus consisting of two kind of nodes, static and mobile nodes. Static Mesh nodes are fixed at a given position forming a grid structure and acts as infrastructure to provide service to the mobile mesh nodes. The static nodes do not generate data traffic. Mobile nodes are placed randomly in the area.

Fig 1. Simulation Scenario showing static and mobile mesh nodes in NS3

3.2.

Parameters

The Simulation scenario consists of 16 static mesh node acting as routers placed 100m apart to provide a coverage area of 500m X 500m and 30 mobile mesh nodes with RandomWayPoint mobility where speed of node is varied from 1 to 10m/s with a fixed pause of 5s. Each node with IEEE 802.11a physical layer and IEEE 80.2.11s based MAC layer is used for simulation. Path selection protocol used is Hybrid wireless mesh protocol (HWMP) with airtime as path selection metric. Each Traffic flows used for simulations consist of UDP packets with randomly initialised payload data of 512 Bytes. Traffic type is Constant bit rate traffic which generate 10 packets per seconds with random start and stops time. The amount of traffic toward root is initially 0% then it is increased with a step of 20% until all traffic is going toward root i.e. 100% traffic toward the root with an additional measurement at 10% traffic rate. All other traffic is send between other nodes which are randomly chosen. Top left static mesh node in the grid is considered as root node, simulating mesh gateway for external traffic. To take reading each simulation was repeated 20 times with different set of random variable to generate sufficient confidence in the results. Four parameters, initial path requests Path selection overhead ratio, path errors and throughput were used to compare behaviour of the mode of HWMP. The path selection overhead ratio is the ratio of bytes sent by the mesh stations as path selection packets to that of all bytes sent by the mesh stations. Throughput, in kbps, is measured as the application level data that is received by static mesh nodes from source mobile mesh nodes and successfully transferred to the sink mesh node. 4. Results Different modes of HWMP experience different path selection overhead as in fig 2. Reactive mode experience least overhead as compared to proactive modes. Among the proactive mode, overhead experienced by forced proactive mode is more than simple proactive mode. Also the path selection overhead also decreases with higher fraction of root traffic.

[42]

Ahmed and Chishti/COMMUNE-2015

Fig 2. Path selection overhead ratio

Fig 3. Number of Path Errors

Fig 4. Number of initial path requests

Fig 5. Throughput at different Hops

[43]

Ahmed and Chishti/COMMUNE-2015

A Path Error packet, PERR, is generated when a path break Path errors decrease as fraction of traffic toward root increases. Proactive mode generates less number of PERRs than Reactive mode. Forced Proactive mode generated least number of PERRs as in fig 3. For a given mode the number of initial path request remain almost constant with a slight decrease as percentage of traffic toward root increases refereeing fig 4. The reactive mode generated least number of path requests whereas proactive mode generated much more number of path requests. The Throughput, fig 5, is evaluated with all traffic sent toward root. Thus, focusing on evaluating throughput experienced by static router at different hops. Irrespective of HWMP mode, throughput decreases as number of hops increases. The Proactive modes show better performance in a root centric environment. Decrease in throughput is more in case of reactive mode. Throughput of simple is less then forced proactive mode because of unidirectional tree and reactive path setup when data is transferred from root to the nodes. 5. Conclusion After analyzing the results author come to following conclusion: Forced proactive mode performs better than simple proactive mode which in turn performs better then reactive mode. As number of hops increases throughput decreases rapidly irrespective of mode of HWMP. Forced proactive mode performs well in the network condition when 100% traffic is directed toward the root node showing higher throughput and least path errors as compared to other modes. However, network must be able to handle high path selection overhead load. In case most of the traffic is outbound and is destined to or from root node, forced proactive mode is a good choice because of better throughput. While analyzing the throughput of the mesh network it is found that the per hop throughput of IEEE 802.11s network do not scale well as number of hops increase. References Perkins, C., Belding-Royer, E., S. Das, S., 2003. AdHoc On-Demand Distance Vector (AODV) Routing, IETF RFC 3561. Chun-Ping Wang, Brett Hagelstein, Mehran Abolhasan, Experimental Evaluation of IEEE 802.11s Path Selection Protocols in a Mesh Testbed, 4th International Conference on Signal Processing and Communication Systems, ICSPCS, IEEE, pp.1-3, 2010. Guido R Hiertz, Sebastian Max, Rakesh Taori, Feb. 2010. IEEE 802.11s: The WLAN Mesh standard, IEEE wireless communications pp. 104-111. Akyildiz I. F., Wang, X., Sept. 2005.A survey on wireless mesh networks, IEEE Commun. Mag, vol. 43, no. 9, pp. S23–S30. IEEE Computer Society, Oct. 2009. IEEE P802.11s/D3.04 Draft Standard: Wireless LAN Medium AccessControl (MAC) and Physical Layer (PHY) specifications, Amendment 10: Mesh Networking. Camp, J. D., Knightly,E.W., August 2008. The IEEE 802.11s Extended Service Set Mesh Networking Standard,IEEE Communications Magazine, vol. 46, no.8, pp. 120–126. Kai Yang, Jian-feng Ma, Zi-hui Miao.,(2009). Hybrid Routing Protocol for Wireless Mesh Network, International Conference on Computational Intelligence and Security. Kuran M.S., Tugcu T., 2007.A survey on emerging broadband wireless access technologies,Comput. Netw., vol. 51, no. 11, pp. 3013–3046. Michael Bahr, 2007. Update on the Hybrid Wireless Mesh Protocol of IEEE 802.11, IEEE 2007 Siemens Corporate Technology, Information & Communications. OpenMesh,Retrived from url: http://www.open-mesh.com. Raffaele Bruno, Marco Conti, Enrico Gregori, March 2005. Mesh Networks: Commodity Multihop Ad Hoc Networks, IEEE Communications Magazine, pp.123-131. Ricardo C. Carrano, Luiz C. S. Magalhães, Débora C. Muchaluat Saade ,2011. Célio V. N. Albuquerque, IEEE 802.11s Multihop MAC: A Tutorial IEEE Communications Surveys & Tutorials, VOL. 13, no. 1, First Quarter. Seungjoon Lee, Suman Banerjee, Bobby Bhattacharjee, 2004.The Case for a Multi-hop Wireless Local Area Network, IEEE.

[44]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16-18March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Ultra Low-Voltage, Robust and Integrable/Programmable Neural Network based Design of 2:1 Multiplexer N. A. Kant*, F. A. Khanday Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar, India

Abstract Hardware implementation of neural networks is one of the contemporary areas of research for the scientists. After a careful survey of the open literature, it is found that apart from some digital implementations and the analog implementation of activation functions/basic logic gates, very less work has been reported for the high-order hardware implementations of neural networks. Digital implementations of neural networks use the sequential machines and therefore loose the basic essence of neural networks. Consequently, there has been an increasing interest in the analog implementations of neural networks. In analog implementation of neural networks, focus is on reducing the supply voltage/power consumption, as the complexity of neural network based circuit is often more than the conventional CMOS based design. Besides the technology is also scaled down to reduce the area of the circuit on the chip. This paper outlines the design and simulated performance of ultra low-voltage multiplexer function designed from the neural network using Sinh domain technique. Two different activation functions are used for the design, where the activation functions are also designed using the Sinh-Domain companding technique. The circuit is designed in 0.35µm AMS CMOS process and the simulation results are obtained from Hspice simulation software.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Artificial Neural Network; Hardware Implementation of Neural Networks; Activation Functions; Analog Integrated Circuits; Companding Techniques; Sinh-Domain Technique; Digital Logic Functions; Multiplexers

1. Introduction Artificial Neural Networks (ANNs) being one of the most widely studied topic for research, now from past few decades, has found its applications in vast variety of fields and is gaining more attention with the advancements in the modern technologies. They have been used for different variety of problems like in the areas of pattern recognition, classification, signal processing, image processing, control systems, biomedical engineering etc. by employing their software based designs. We have many examples in open literature showing neural network codes running on von Neumann computers. In spite of many years of studies involving Neural Networks (NNs), their actual implementation for study and their applications have become practical owing to the advancement in programmable hardware of these networks. As the key features of neural networks are parallel processing, continuous-time dynamics, and global interaction of network elements [Hopfield, 1982] and all these can be fully utilized only by its hardware design rather than software approach. Therefore, it is important to look out for the methods by which one can achieve the hardware design of NNs. But then again also to mention that the hardware designs of these networks is going at a dawdling pace and there are still few commercially available neural networks implemented in hardware [Chao-Ming eta al., 2007, Seul and Sung, 2007]. The biggest concern while designing the neural network is the complexity and their training. As the complexity increases the concern shifts towards the power consumption and the voltage of operation. In the aforementioned context, flexibility and power consumption (in order to satisfy a wide range of applications) are the important aspects to be kept into consideration in the hardware implementation of neural networks. The designer must thus look for the design technique to achieve the said goal of low-power low-voltage hardware circuit for neural network. *

Corresponding author. Tel.: +91 9906 937561. E-mail address: [email protected].

ISBN: 978-93-82288-63-3

Khanday and Kant/COMMUNE – 2015

Hardware implementation of ANNs can be achieved through analog or digital means. However, analog realization of the NNs provides a fast and power efficient realization compared to the digital realization whereas the feature of flexibility is added by employing tunable analog-hardware. Now to achieve low-power low-voltage hardware design of these networks companding design is a good choice. Companding design techniques are interesting subclass of analog design with potential for low-voltage operation capability. This is originated from the fact that it is the compressed signals that are processed, which results in the reduction of swing compared with that of the signals in the conventional linear circuit design. NNs can be used to solve any mathematical/engineering problem, multiplexer being one of them. Multiplexer, frequently termed as MUX, an important logic building block which find its application in almost all the digital integrated circuits. MUX find its application in the field of telecommunications and are extremely important, as they are very helpful in reducing network complexity by minimizing the number of communications links needed between two points and in turn reducing the cost. Along all the computing systems, Multiplexer have evolved with time. Each new generation has additional intelligence, and additional intelligence brings more benefits. Few of the accrued benefits are [www.informit.com] a) b) c)

Data compression: The capability to do data compression, which enables us to encode certain characters with smaller number of bits than normally required. The freed capacity can be utilized for movement of other information. Error detection and correction: Data integrity and accuracy are maintained by error detection and correction between the two points that have being connected. Managing transmission resources: The capability to manage transmission resources on a dynamic basis, by introducing the additional features such as priority levels.

Device cost versus line cost is an important issue in communication for transferring signal from one point to the other. We can either provide extremely high levels of service by ensuring everybody always has a continuous and live communications link. This however has a drawback of extremely high cost. The other option we can opt to offset the costs associated with providing large numbers of lines is by instead using devices such as multiplexers that helps in making more intelligent use of a smaller group of lines. The more intelligent the multiplexer, the more perceptively and vigorously it can work on our behalf to dynamically make use of the available transmission resources. In this paper, first time in the literature, a NN based 2:1 MUX design is reported. This has been achieved by employing Sinh-Domain technique. The design has been implemented using two activation functions and important performance parameters have been calculated. The paper is organized as follows: the brief idea about Sinh-Domain companding along with the NN is presented in Section 2. The circuit description of 2:1 MUX along with the simulated results is presented in Section 3. The paper is finally concluded by section 4 and references. 2. Sinh-Domain Companding Technique and NN Design There are a number of techniques which have been used in the literature to achieve the low-voltage, low-power circuit designs. Companding being a technique which is the preferred one for low voltage, low power circuit design due to its inherent advantages of low-voltage requirement, grounded passive component requirement, class AB behavior, electronic tunability etc.. Sinh-Domain companding which is a sub class of companding technique comes in the category of instantaneous companding where the compression of the input current could be performed through the inverse of the hyperbolic sine function realized by translinear loops formed by bipolar transistors in active region or MOS transistors in weak inversion [Serdijn et al., 1999, Katsiamis et al., 2008, Frey, 1996]. The non-linear transconductor cell, which forms the basic building block of sinh-domain technique, is depicted in Fig. 1. This circuit realize the expression given in (1) in the case of the hyperbolic sine output

 vˆ  vˆ IN  i  2 I o sinh IN   nVT

  

(1)

 vˆ  vˆ  i  2 I o cosh IN  IN    nVT 

(2)

(2) in the case for hyperbolic cosine output

(3) in the case for inverted hyperbolic sine output  vˆ  vˆ  i  2 I o sinh IN  IN    nVT 

and (4) in the case for weighted hyperbolic sine output

[46]

(3)

Khanday and Kant/COMMUNE – 2015

 vˆ  vˆ  i  2 KI o sinh IN  IN    nVT 

(4)

where Io is a dc current, VT is the thermal voltage (26mV @ 27oC), n is the subthreshold slope factor (1do)

obj

agt

Seema(icl>person)

Fig. 1. UNL representation of ‫ِکرکٹ‬

@entry.@present.@progress

Cricket (icl>game)

‫ندان‬ٛ ‫چھے ِگ‬ ‫سیما‬ ٚ ٖ

In this graph “Play (icl >do)”, Seema (icl>person) and Cricket (icl>game) are UW’s . agt (agent) and obj (object) are relations and @entry.@present.@progress are attributes. UNL Expression: [UNL] agt(play(icl>do)@entry.@present.@progress, seema(iof>person)) obj(play(icl>do)@entry.@present.@progress,cricket(icl>game)) [/UNL] 3.

A Brief Description of UNL system

UNL system consists of UNL (which involves UW’s, Relations and Attributes), Language servers (enconverter and Deconverter) and Basic tools. With access to the internet one can convert NL to UNL and viceversa. 4.

UNLisation Framework of Kashmiri phrases

Enconverter is the core software in the UNL system. Enonverter converts a given sentence in natural language to an equivalent UNL expression. Enonverter converts a given sentence to UNL expression by accessing word dictionary and interpreting Analysis Rules. The Enconverter transforms all natural language sentences to UNL expressions using a dictionary and a set of grammar rules of the respective language. Enconverter is a language independent parser which can deal with various languages using respective dictionaries and enconversion rules. ENCONVERTER

NL

Universal word Dictionary

UNL EXPRESSION

UNL kB

Analysis Rules

Fig. 1. Components of Enconverter

[116]

Kak et al/COMMUNE – 2015

Using a NL corpus

Formulation of Kashmiri Dictionary in IAN

Rules for enconversion of Kashmiri corpus to UNL in IAN

To any language with its own De converter

UNL

Fig. 2. Scheme of Enconversion

The format of entries in a word dictionary is as: [HW]{ID}”UW”(ATTR….) Uchida, H. and M. Zhu, 1998. “The Universal Networking Language (UNL) Specification Version 3.0”, Technical Report, United Nations University, Tokyo. Uchida, H., M. Zhu and T.C.D. Senta, 2005. Universal Networking Language, UNDL Foundation: Geneva. UNLweb. 2010. UNDL Foundation.

[118]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Estimation of Stress-Strength Reliability Using Finite Mixture of Exponential and Gamma Distributions Adil H. Khan*, T. R. Jan Department of Statistics, University of Kashmir, Srinagar, India

Abstract The term “stress- strength reliability” refers to the quantity 𝑃(𝑋 > 𝑌), where a system with random strength X is subjected to a random stress Y such that a system fails, if the stress exceeds the strength. In this paper Stress –Strength reliability is considered where various cases have been considered for stress (Y) and strength (X) variables viz., the strength follows finite mixture of one parameter exponential and one, two parameter Gamma distributions and stress follows one and two parameter exponential distribution. The general expressions for the reliabilities of a system are obtained. Special cases are also discussed. At the end, results are illustrated with the help of numerical evaluations.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Exponential Distribution; Gamma Distribution; Reliability; Stress; Strength

1. Introduction Now a day’s industrial world has daily been facing with new systems required high level of safety. In this regard, great attentions have been attracted by safety issues in the recent years. It is reported that system importance has a direct relationship with system safety. Thus, there is an increasing need to work on system safety leading reliability theory. There are appliances (every physical component possesses an inherent strength) which survive due to their strength. These appliances receive a certain level of stress and sustain. However, if a higher level of stress is applied then their strength is unable to sustain and they break down. Suppose Y represents the ‘stress’ which is applied to a certain appliance and X represents the ‘strength’ to sustain the stress, then the stress-strength reliability is denoted by R= P(Y 0, 𝜆 > 0, 𝑝𝑖 > 0 𝑖 = 1,2, … , 𝑘 𝑎𝑛𝑑 ∑ 𝑝𝑖 = 1 𝑖=1

[120]

Khan and Jan /COMMUNE – 2015

𝑔(𝑦, 𝛼, 𝜆) = 𝜆𝑒 −𝜆(𝑦−𝛼) ∶ 𝑦 > 𝛼 ≥ 0, 𝜆 > 0 For two components k = 2, we have 𝑓(𝑥, 𝜆1 , 𝜆2 ) = 𝑝1 𝜆1 𝑒 𝜆1𝑥 + 𝑝2 𝜆2 𝑒 𝜆2𝑥 ; 𝑥, 𝜆, 𝑝1 , 𝑝2 > 0, In addition, if X and Y are independent, then from (2.1) reliability R is given by

𝑝1 + 𝑝2 = 1

∞ 𝑥

𝑅2 = ∫ ∫(𝜆𝑒 −𝜆(𝑦−𝛼) )(𝑝1 𝜆1 𝑒 𝜆1𝑥 + 𝑝2 𝜆2 𝑒 𝜆2𝑥 ) 𝑑𝑥𝑑𝑦 0 𝛼 2



𝑅2 = ∑ 𝑝𝑖 𝜆𝑖 [∫ 𝑒 𝜆𝑖𝑥 (1 − 𝑒 −𝜆(𝑥−𝛼) )𝑑𝑥 ] 𝑖=1

𝛼

2

𝑅2 = 𝜆 ∑ 𝑝𝑖 𝑖=1

𝑒 −𝛼𝜆𝑖 𝜆 + 𝜆𝑖

For three components k = 3, we have𝑝1 𝜆1 𝑒 𝜆1𝑥 + 𝑝2 𝜆2 𝑒 𝜆2𝑥 + 𝑝3 𝜆3 𝑒 𝜆3𝑥 𝑓(𝑥, 𝜆1 , 𝜆2 , 𝜆3 ) = 𝑝1 𝜆1 𝑒 𝜆1𝑥 + 𝑝2 𝜆2 𝑒 𝜆2𝑥 + 𝑝3 𝜆3 𝑒 𝜆3𝑥 ; 𝑥, 𝜆, 𝑝1 , 𝑝2 , 𝑝3 > 0, 𝑝1 + 𝑝2 + 𝑝3 = 1 In addition, if X and Y are independent, then from (1) reliability R is given by ∞ 𝑥

𝑅3 = ∫ ∫(𝜆𝑒 −𝜆(𝑦−𝛼) )(𝑝1 𝜆1 𝑒 𝜆1𝑥 + 𝑝2 𝜆2 𝑒 𝜆2𝑥 + 𝑝3 𝜆3 𝑒 𝜆3𝑥 ) 𝑑𝑥𝑑𝑦 0 𝛼

3



𝑅3 = ∑ 𝑝𝑖 𝜆𝑖 [∫ 𝑒 𝜆𝑖𝑥 (1 − 𝑒 −𝜆(𝑥−𝛼) )𝑑𝑥 ] 𝑖=1

𝛼

3

𝑅3 = 𝜆 ∑ 𝑝𝑖 𝑖=1

𝑒 −𝛼𝜆𝑖 𝜆 + 𝜆𝑖

In general, for k-components, we have 𝑓(𝑥, 𝜆1 , 𝜆2 , … , 𝜆𝑘 ) = 𝑝1 𝜆1 𝑒 𝜆1𝑥 + 𝑝2 𝜆2 𝑒 𝜆2𝑥 + ⋯ + 𝑝𝑘 𝜆𝑘 𝑒 𝜆𝑘𝑥 ; 𝑥, 𝜆, 𝑝1 , 𝑝2 , … , 𝑝𝑘 > 0, 𝑝1 + 𝑝2 + ⋯ + 𝑝𝑘 = 1 𝑘

and

𝑅𝑘 = 𝜆 ∑ 𝑝𝑖 𝑖=1

𝑒 −𝛼𝜆𝑖 𝜆 + 𝜆𝑖

Special case: When 𝛼 = 0, two-parameter exponential distribution reduces to one parameter distribution. Therefore, 𝑅𝑘 when strength X follows mixture of exponential distribution and stress Y follows exponential distribution is given by 𝑘

𝑘

𝑖=1

𝑖=1

1 𝑝𝑖 𝜆𝑖 𝑅𝑘 = 𝜆 ∑ 𝑝𝑖 =1−∑ 𝜆 + 𝜆𝑖 𝜆 + 𝜆𝑖 (See (Umamaheswan, Sandhya, 2013)) Case II: Mixture of one parameter gamma strength and exponential stress. Let X be the strength of k-components which follows mixture of one parameter gamma distribution with pdf 𝑓𝑖 (𝑥, 𝛽𝑖 ) and Y be the stress which follows exponential distribution with pdf 𝑔(𝑦, 𝜆), where 𝑘

𝑒 −𝑥 𝑥 𝛽𝑖−1 𝑓𝑖 (𝑥, 𝛽𝑖 ) = 𝑝𝑖 ; 𝑥 > 0, 𝛽𝑖 > 0, 𝑝𝑖 > 0 𝑖 = 1,2, … , 𝑘 𝑎𝑛𝑑 ∑ 𝑝𝑖 = 1 Γ(𝛽𝑖 ) 𝑔(𝑦, 𝜆) = 𝜆𝑒 −𝜆𝑦 ∶ 𝑦 > 0, 𝜆 > 0 For two components k = 2, we have 𝑒 −𝑥 𝑥 𝛽1 −1 𝑒 −𝑥 𝑥 𝛽2 −1 𝑓(𝑥, 𝛽1 , 𝛽2 ) = 𝑝1 + 𝑝2 ; 𝑝1 + 𝑝2 = 1 Γ(𝛽1 ) Γ(𝛽2 ) In addition, if X and Y are independent, then from (2.1) reliability R is given by ∞ 𝑥

𝑅2 = ∫ ∫(𝜆𝑒 −𝜆𝑦 ) (𝑝1 0 0

𝑒 −𝑥 𝑥 𝛽1−1 𝑒 −𝑥 𝑥 𝛽2−1 + 𝑝2 ) 𝑑𝑥𝑑𝑦 Γ(𝛽1 ) Γ(𝛽2 )

[121]

𝑖=1

(3.1)

Khan and Jan /COMMUNE – 2015 ∞

2

𝑝𝑖 𝑅2 = ∑ [∫(1 − 𝑒 −𝜆𝑥 )𝑒 −𝑥 𝑥 𝛽𝑖 −1 𝑑𝑥 ] Γ(𝛽𝑖 ) 𝑖=1

0

2

𝑝𝑖 (1 + 𝜆)𝛽𝑖

𝑅2 = 1 − ∑ 𝑖=1

For three components k = 3, we have 𝑓(𝑥, 𝛽1 , 𝛽2 , 𝛽3 ) = 𝑝1

𝑒 −𝑥 𝑥 𝛽1−1 𝑒 −𝑥 𝑥 𝛽2−1 𝑒 −𝑥 𝑥 𝛽3−1 + 𝑝2 + 𝑝3 ; 𝑝1 + 𝑝2 + 𝑝3 = 1 Γ(𝛽1 ) Γ(𝛽2 ) Γ(𝛽3 )

In addition, if X and Y are independent, then from (2.1) reliability R is given by ∞ 𝑥

𝑅3 = ∫ ∫(𝜆𝑒 −𝜆𝑦 ) (𝑝1 0 0

𝑒 −𝑥 𝑥 𝛽1−1 𝑒 −𝑥 𝑥 𝛽1−1 𝑒 −𝑥 𝑥 𝛽3−1 + 𝑝2 + 𝑝3 ) 𝑑𝑥𝑑𝑦 Γ(𝛽1 ) Γ(𝛽1 ) Γ(𝛽3 ) ∞

3

𝑝𝑖 𝑅3 = ∑ [∫(1 − 𝑒 −𝜆𝑥 )𝑒 −𝑥 𝑥 𝛽𝑖 −1 𝑑𝑥 ] Γ(𝛽𝑖 ) 𝑖=1

0

3

𝑝𝑖 (1 + 𝜆)𝛽𝑖

𝑅3 = 1 − ∑ 𝑖=1

In general, for k-components, we have 𝑒 −𝑥 𝑥 𝛽1−1 𝑒 −𝑥 𝑥 𝛽2−1 𝑒 −𝑥 𝑥 𝛽𝑘 −1 𝑓(𝑥, 𝛽1 , 𝛽2 , … , 𝛽𝑘 ) = 𝑝1 + 𝑝2 + ⋯ + 𝑝𝑘 ; 𝑝1 + 𝑝2 + ⋯ + 𝑝𝑘 = 1 Γ(𝛽1 ) Γ(𝛽2 ) Γ(𝛽𝑘 ) 𝑘

and

𝑅𝑘 = 1 − ∑ 𝑖=1

𝑝𝑖 (1 + 𝜆)𝛽𝑖

Case III: Mixture of two parameter gamma strength and exponential stress. Let X be the strength of k-components which follows mixture of two parameter gamma distribution with pdf 𝑓𝑖 (𝑥, 𝜆𝑖 , 𝛽𝑖 ) and Y be the stress which follows two parameter exponential distribution with pdf (3.1), where 𝑘

𝜆𝑖 𝛽𝑖 𝑒 −𝜆𝑖𝑥 𝑥 𝛽𝑖 −1 𝑓𝑖 (𝑥, 𝜆𝑖 , 𝛽𝑖 ) = 𝑝𝑖 ; 𝑥 > 0, 𝜆𝑖 , 𝛽𝑖 , 𝑝𝑖 > 0 𝑖 = 1,2, … , 𝑘 𝑎𝑛𝑑 ∑ 𝑝𝑖 = 1 Γ(𝛽𝑖 ) 𝑖=1

For two components k = 2, we have 𝑓(𝑥, 𝜆1 , 𝜆2 , 𝛽1 , 𝛽2 ) = 𝑝1

𝜆1 𝛽1 𝑒 −𝜆1 𝑥 𝑥 𝛽1−1 𝜆2 𝛽2 𝑒 −𝜆2𝑥 𝑥 𝛽2 −1 + 𝑝2 ; 𝑝1 + 𝑝2 = 1 Γ(𝛽1 ) Γ(𝛽2 )

In addition, if X and Y are independent, then from (1) reliability R is given by ∞ 𝑥

𝑅2 = ∫ ∫(𝜆𝑒 −𝜆𝑦 ) (𝑝1 0 0

𝜆1 𝛽1 𝑒 −𝜆1 𝑥 𝑥 𝛽1−1 𝜆2 𝛽2 𝑒 −𝜆2𝑥 𝑥 𝛽2 −1 + 𝑝2 ) 𝑑𝑥𝑑𝑦 Γ(𝛽1 ) Γ(𝛽2 ) ∞

2

𝑝𝑖 𝜆𝑖 𝛽𝑖 𝑅2 = ∑ [∫ (1 − 𝑒 −𝜆𝑥 )𝑒 −𝜆𝑖𝑥 𝑥 𝛽𝑖−1 𝑑𝑥 ] Γ(𝛽𝑖 ) 𝑖=1

0

2

𝑅2 = 1 − ∑ 𝑖=1

𝑝𝑖 𝜆𝑖 𝛽𝑖 (𝜆 + 𝜆𝑖 )𝛽𝑖

For three components k = 3, we have

𝜆1 𝛽1 𝑒 −𝜆1𝑥 𝑥 𝛽1−1 𝜆2 𝛽2 𝑒 −𝜆2𝑥 𝑥 𝛽2 −1 𝜆3 3 𝑒 −𝜆3 𝑥 𝑥 𝛽3−1 + 𝑝2 + 𝑝3 Γ(𝛽1 ) Γ(𝛽2 ) Γ(𝛽3 ) ; 𝑝1 + 𝑝2 + 𝑝3 = 1 In addition, if X and Y are independent, then from (1) reliability R is given by 𝑓(𝑥, 𝜆1 , 𝜆2 , 𝜆3 , 𝛽1 , 𝛽2 , 𝛽3 ) = 𝑝1

∞ 𝑥

𝑅3 = ∫ ∫(𝜆𝑒 −𝜆𝑦 ) (𝑝1 0 0

𝜆1 𝛽1 𝑒 −𝜆1𝑥 𝑥 𝛽1−1 𝜆2 𝛽2 𝑒 −𝜆2𝑥 𝑥 𝛽2 −1 𝜆3 3 𝑒 −𝜆3𝑥 𝑥 𝛽3−1 + 𝑝2 + 𝑝3 ) 𝑑𝑥𝑑𝑦 Γ(𝛽1 ) Γ(𝛽2 ) Γ(𝛽3 ) 3



𝑝𝑖 𝜆𝑖 𝛽𝑖 𝑅3 = ∑ [∫ (1 − 𝑒 −𝜆𝑥 )𝑒 −𝜆𝑖𝑥 𝑥 𝛽𝑖−1 𝑑𝑥 ] Γ(𝛽𝑖 ) 𝑖=1

0

[122]

Khan and Jan /COMMUNE – 2015 3

𝑅3 = 1 − ∑ 𝑖=1

In general, for k-components, we have 𝑓(𝑥, 𝜆1 , … , 𝜆𝑘 , 𝛽1 , … , 𝛽𝑘 ) =

𝑝1 𝜆1 𝛽1 𝑒 −𝜆1 𝑥 𝑥 𝛽1−1 𝑝2 𝜆2 𝛽2 𝑒 −𝜆2𝑥 𝑥 𝛽2−1 𝑝𝑘 𝜆𝑘 𝑘 𝑒 −𝜆𝑘 𝑥 𝑥 𝛽𝑘 −1 + + ⋯+ ; Γ(𝛽1 ) Γ(𝛽2 ) Γ(𝛽𝑘 ) 𝑝1 + 𝑝2 + ⋯ + 𝑝𝑘 = 1 𝑘

and

𝑝𝑖 𝜆𝑖 𝛽𝑖 (𝜆 + 𝜆𝑖 )𝛽𝑖

𝑅𝑘 = 1 − ∑ 𝑖=1

𝑝𝑖 𝜆𝑖 𝛽𝑖 (𝜆 + 𝜆𝑖 )𝛽𝑖

Case IV: Mixture of two-parameter gamma strength and mixture of exponential stress. Let X be the strength of k-components which follows mixture of two parameter gamma distribution with pdf 𝑓𝑖 (𝑥, 𝜆𝑖 , 𝛽𝑖 ) and Y be the stress which follows two parameter exponential distribution with pdf 𝑔(𝑦, 𝜆𝑗 ), where 𝑓𝑗 (𝑥, 𝜆𝑗 , 𝛽𝑗 ) = 𝑝𝑗

𝜆𝑗 𝛽𝑗 𝑒 −𝜆𝑗 𝑥 𝑥 𝑗−1 Γ(𝛽𝑗 )

2𝑘

; 𝑥 > 0, 𝜆𝑗 , 𝛽𝑗 , 𝑝𝑗 > 0 𝑗 = 1,2, … , 𝑘 𝑎𝑛𝑑 ∑ 𝑝𝑗 = 1 𝑘

𝑗=𝑘+1

𝑔𝑖 (𝑦, 𝜆𝑖 ) = 𝑝𝑖 𝜆𝑖 𝑒 −𝜆𝑖𝑦 ∶ 𝑥 > 0, 𝜆𝑖 > 0 𝑎𝑛𝑑 ∑ 𝑝𝑖 = 1 𝑖=1

For two components k = 2, we have 𝑔(𝑦, 𝜆1 , 𝜆2 ) = 𝑝1 𝜆1 𝑒 −𝜆1𝑦 + 𝑝2 𝜆2 𝑒 −𝜆2𝑦 ; 𝑝1 + 𝑝2 = 1 𝑓(𝑥, 𝜆3 , 𝜆4 , 𝛽3 , 𝛽4 ) = 𝑝3

𝜆3 𝛽3 𝑒 −𝜆3 𝑥 𝑥 𝛽3−1 𝜆4 𝛽4 𝑒 −𝜆4𝑥 𝑥 𝛽4−1 + 𝑝4 ; 𝑝3 + 𝑝4 = 1 Γ(𝛽3 ) Γ(𝛽4 )

In addition, if X and Y are independent, then from (1) reliability R is given by ∞ 𝑥

𝑅2 = ∫ ∫(𝑝1 𝜆1 𝑒 −𝜆1𝑦 + 𝑝2 𝜆2 𝑒 −𝜆2𝑦 ) (𝑝3 0 𝛼

4

2

𝑅2 = ∑ ∑ 𝑗=𝑖+2 𝑖=1

𝑝𝑖 𝑝𝑗 𝜆𝑗 𝛽𝑗 Γ(𝛽𝑗 ) 4

𝜆3 𝛽3 𝑒 −𝜆3𝑥 𝑥 𝛽3 −1 𝜆4 𝛽4 𝑒 −𝜆4 𝑥 𝑥 𝛽4−1 + 𝑝4 ) 𝑑𝑥𝑑𝑦 Γ(𝛽3 ) Γ(𝛽4 )



[∫ (1 − 𝜆𝑖 𝑒 −𝜆𝑖𝑥 )𝑒 −𝜆𝑗 𝑥 𝑥 𝛽𝑗 −1 𝑑𝑥 ] 0

2

𝑅2 = 1 − ∑ ∑ 𝑝𝑖 𝑝𝑗 𝑗=𝑖+2 𝑖=1

𝜆𝑗 𝛽𝑗 𝛽𝑗

(𝜆𝑖 + 𝜆𝑗 )

For three components k = 3, we have 𝑔(𝑥, 𝜆1 𝜆2 , 𝜆3 , ) = 𝑝1 𝜆1 𝑒 −𝜆1𝑦 + 𝑝2 𝜆2 𝑒 −𝜆2𝑦 + 𝑝3 𝜆3 𝑒 −𝜆3𝑦 ;

𝑝1 + 𝑝2 + 𝑝2 = 1 𝜆3 𝛽3 𝑒 −𝜆3𝑥 𝑥 𝛽3−1 𝜆4 𝛽4 𝑒 −𝜆4𝑥 𝑥 𝛽4−1 𝜆5 𝛽5 𝑒 −𝜆5𝑥 𝑥 𝛽5−1 𝑓(𝑥, 𝜆3 , 𝜆4 , 𝜆5 , 𝛽3 , 𝛽4 , 𝛽5 ) = 𝑝3 + 𝑝4 + 𝑝5 ; Γ(𝛽3 ) Γ(𝛽4 ) Γ(𝛽5 ) 𝑝3 + 𝑝4 + 𝑝5 = 1 In addition, if X and Y are independent, then from (1) reliability R is given by ∞ 𝑥

𝑅3 = ∫ ∫(𝑝1 𝜆1 𝑒 −𝜆1𝑦 + 𝑝2 𝜆2 𝑒 −𝜆2𝑦 + 𝑝3 𝜆3 𝑒 −𝜆3𝑦 ) (𝑝3 0 0

+ 𝑝5

𝜆3 𝛽3 𝑒 −𝜆3 𝑥 𝑥 𝛽3−1 𝜆4 𝛽4 𝑒 −𝜆4𝑥 𝑥 𝛽4−1 + 𝑝4 Γ(𝛽3 ) Γ(𝛽4 )

𝜆5 𝛽5 𝑒 −𝜆5𝑥 𝑥 𝛽5 −1 ) 𝑑𝑥𝑑𝑦 Γ(𝛽5 ) 6

3

𝑅3 = ∑ ∑ 𝑗=𝑖+2 𝑖=1

𝑝𝑖 𝑝𝑗 𝜆𝑗 𝛽𝑗 Γ(𝛽𝑗 ) 6



[∫ (1 − 𝜆𝑖 𝑒 −𝜆𝑖𝑥 )𝑒 −𝜆𝑗 𝑥 𝑥 𝛽𝑗 −1 𝑑𝑥 ] 0

3

𝑅3 = 1 − ∑ ∑ 𝑝𝑖 𝑝𝑗 𝑗=𝑖+2 𝑖=1

In general, for k-components, we have

[123]

𝜆𝑗 𝛽𝑗 𝛽𝑗

(𝜆𝑖 + 𝜆𝑗 )

Khan and Jan /COMMUNE – 2015

𝑔(𝑥, 𝜆1 , … , 𝜆𝑘 ) = 𝑝1 𝜆1 𝑒 −𝜆1𝑦 + 𝑝2 𝜆2 𝑒 −𝜆2𝑦 + ⋯ + 𝑝𝑘 𝜆𝑘 𝑒 −𝑘𝑦 ;

𝑝1 + 𝑝2 + ⋯ + 𝑝𝐾 = 1 𝑓(𝑥, 𝜆𝑘+1 , … , 𝜆2𝑘 , 𝛽𝑘+1 , … , 𝛽2𝑘 ) 𝜆𝑘+1 𝛽𝑘+1 𝑒 −𝜆𝑘+1𝑥 𝑥 𝛽𝑘+1−1 𝜆𝑘+2 𝛽𝑘+2 𝑒 −𝜆𝑘+2 𝑥 𝑥 𝛽𝑘+2−1 = 𝑝𝑘+1 + 𝑝𝑘+2 +⋯ Γ(𝛽𝑘+1 ) Γ(𝛽𝑘+2 ) 𝜆2𝑘 𝛽2𝑘 𝑒 −𝜆2𝑘 𝑥 𝑥 𝛽2𝑘 −1 + 𝑝2𝑘 ; 𝑝𝑘+1 + 𝑝𝑘+2 + ⋯ + 𝑝2𝑘 = 1 Γ(𝛽2𝑘 ) ∞ 𝑥

𝑅𝑘 = ∫ ∫(𝑝1 𝜆1 𝑒 −𝜆1𝑦 + 𝑝2 𝜆2 𝑒 −𝜆2 𝑦 + ⋯ + 𝑝𝑘 𝜆𝑘 𝑒 −𝑘𝑦 ) (𝑝𝑘+1 0 0

+ 𝑝𝑘+2

𝜆𝑘+1 𝛽𝑘+1 𝑒 −𝜆𝑘+1𝑥 𝑥 𝛽𝑘+1−1 Γ(𝛽𝑘+1 )

𝜆𝑘+2 𝛽𝑘+2 𝑒 −𝜆𝑘+2 𝑥 𝑥 𝛽𝑘+2−1 𝜆2𝑘 𝛽2𝑘 𝑒 −𝜆2𝑘 𝑥 𝑥 𝛽2𝑘 −1 + ⋯ + 𝑝2𝑘 ) 𝑑𝑥𝑑𝑦 Γ(𝛽𝑘+2 ) Γ(𝛽2𝑘 ) 2𝑘

𝑘

𝑅𝑘 = 1 − ∑ ∑ 𝑝𝑖 𝑝𝑗 𝑗=𝑖+2 𝑖=1

𝜆𝑗 𝛽𝑗 (𝜆𝑖 + 𝜆𝑗 )

𝛽𝑗

4. Numerical Evaluation For some specific values of the parameters involved in the expression 𝑅𝑘 (𝑘 = 2), we have evaluated the system reliability for different cases of Exponential, Gamma distributions from their expression obtained in case 1, 2, 3 and 4. Mixture of one parameter exponential strength and two parameter exponential stress Table 1 𝜆 0.8 0.8 0.8 0.8 0.8 0.8 0.8

𝜆1 = 𝜆2 0 0.2 0.4 0.6 0.8 1 1.2

𝛼 0.2 0.2 0.2 0.2 0.2 0.2 0.2

Table 2 𝑅 1 0.7686 0.6154 0.5068 0.4260 0.3638 0.3146

𝜆 0.8 0.8 0.8 0.8 0.8 0.8 0.8

𝜆1 = 𝜆2 0.3 0.3 0.3 0.3 0.3 0.3 0.3

Table 3

𝛼 0 0.2 0.4 0.6 0.8 1 1.2

𝑅 0.7272 0.6849 0.6450 0.6074 0.5720 0.5387 0.5074

𝜆 0 0.2 0.4 0.6 0.8 1 1.2

𝜆1 = 𝜆2 0.3 0.3 0.3 0.3 0.3 0.3 0.3

𝛼 0.2 0.2 0.2 0.2 0.2 0.2 0.2

𝑅 0 0.3767 0.5381 0.6278 0.6849 0.7244 0.7534

Mixture of one parameter gamma strength and exponential stress Table 4

Table 5

𝝀

𝜷𝟏 = 𝜷𝟐

𝑹

𝝀

𝜷𝟏 = 𝜷𝟐

𝑹

0.7 0.7 0.7 0.7 0.7 0.7 0.7

0 0.2 0.4 0.6 0.8 1 1.2

0 0.1006 0.1912 0.2726 0.3459 0.4117 0.4709

0 0.2 0.4 0.6 0.8 1 1.2

0.7 0.7 0.7 0.7 0.7 0.7 0.7

0 0.1198 0.2098 0.2803 0.3373 0.3844 0.4241

Mixture of two parameter gamma strength and exponential stress Table 6

Table 7

Table 8

𝝀

𝝀𝟏 = 𝝀𝟐

𝜷𝟏 = 𝜷𝟐

𝑹

𝝀

𝝀𝟏 = 𝝀𝟐

𝜷𝟏 = 𝜷𝟐

𝑹

𝝀

𝝀𝟏 = 𝝀𝟐

𝜷𝟏 = 𝜷𝟐

𝑹

0 0.2 0.4 0.6 0.8 1 1.2

0.4 0.4 0.4 0.4 0.4 0.4 0.4

0.7 0.7 0.7 0.7 0.7 0.7 0.7

0 0.2471 0.3844 0.4734 0.5365 0.5839 0.6210

0.3 0.3 0.3 0.3 0.3 0.3 0.3

0.6 0.6 0.6 0.6 0.6 0.6 0.6

0 0.2 0.4 0.6 0.8 1 1.2

0 0.0778 0.1497 0.2159 0.2770 0.3333 0.3852

0.3 0.3 0.3 0.3 0.3 0.3 0.3

0 0.2 0.4 0.6 0.8 1 1.2

0.5 0.5 0.5 0.5 0.5 0.5 0.5

1 0.3675 0.2440 0.1835 0.1471 0.1229 0.1055

[124]

Khan and Jan /COMMUNE – 2015

Mixture of two parameter gamma strength and mixture of exponential stress Table 9

Table 10

Table 11

𝝀𝟏 = 𝝀𝟐

𝝀𝟑 = 𝝀𝟒

𝜷𝟑 = 𝜷𝟒

𝑹

𝝀𝟏 = 𝝀𝟐

𝝀𝟑 = 𝝀𝟒

𝜷𝟑 = 𝜷𝟒

𝑹

𝝀𝟏 = 𝝀𝟐

𝝀𝟑 = 𝝀𝟒

𝜷𝟑 = 𝜷𝟒

𝑹

0 0.2 0.4 0.6 0.8 1 1.2

0.2 0.2 0.2 0.2 0.2 0.2 0.2

0.6 0.6 0.6 0.6 0.6 0.6 0.6

0 0.3402 0.4827 0.5647 0.6192 0.6587 0.6888

0.2 0.2 0.2 0.2 0.2 0.2 0.2

0 0.2 0.4 0.6 0.8 1 1.2

0.6 0.6 0.6 0.6 0.6 0.6 0.6

1 0.3402 0.2159 0.1585 0.1253 0.1036 0.0883

0.2 0.2 0.2 0.2 0.2 0.2 0.2

0.6 0.6 0.6 0.6 0.6 0.6 0.6

0 0.2 0.4 0.6 0.8 1 1.2

0 0.0559 0.1086 0.1585 0.2055 0.2500 0.2919

5. Conclusion In this paper Stress –Strength reliability is considered where various cases have been considered for stress (Y) and strength (X) variables viz., the strength follows finite mixture of one parameter exponential and one, two parameter Gamma distributions and stress follows one and two parameter exponential distribution. From Table 1, it may be observed that reliability decreases when both the stress parameters are kept constant and strength variable is increased, in table 2, reliability increases if one of the stress variable 𝛼 is increased (𝜆 fixed) and strength variables 𝜆1 , 𝜆2 are constants. Similarly, from table 3, reliability increases with increase in stress variable 𝜆 (𝛼 constant) and strength being constant. Table 4 and Table 5 exhibits that reliability increases in both the cases i.e. if we increase the stress or strength variables reliability increases. For example, in Table 4 if we increase the strength variables from 0 upto 1.2 the reliability increases from 0 upto 0.4709. Table 6 reveals that if we increase the stress and keep the strength constant reliability increases from 0 upto 0.6210. Again, with increase in one of the strength variable (𝛽1 = 𝛽2 ) keeping stress constant, reliability increases, but in table 8 reliability decreases when stress is constant and strength variable (𝜆1 = 𝜆2 ) is increased. Table 9 exhibits that if we increase the stress and keep strength constant them the reliability of the system increases. Similarly, in table 11 reliability increases but this time stress is constant and one of the strength variables 𝛽3 = 𝛽4 (𝜆3 = 𝜆4 , constant). And in table 10 reliability decreases with increase in strength variable and keeping stress constant, for example, if we increase 𝜆3 = 𝜆4 from 0 upto 1.2 reliability decreases from 1 upto 0.0883. References Awad, A. M. and Gharraf, M. K., 1986. Estimation of (P(Y < X) in the Burr case: A comparative study, Commun. Statist. Simul. Comp., 15(2), p-389403. Beg, M. A.and Singh, N., 1979. Estimation of P(Y < X) for the pareto distribution, IEEE Trans. Reliab., 28(5), p- 411-414. Chaturvadi, A., Tiwari, N. and Kumar, S., 2007. Some remarks on classical and Bayesian reliability estimation of binomial and Poisson distributions, Statistical papers, 48, p-683-693. Chaturvedi, A., and Tomer, S. K., 2002. Classical and Bayesian Reliability estimation of the negative binomial distribution, Jour. Applied Statist. Sci., 11(1), p-33-43. Church, J. D. and Harris, B., 1970. The estimation of reliability from stress strength relationships, Technometrics, 12, p-49-54. Gogoi J., and Borah, M., 2012. Estimation of Reliability for Multi-Component Systems Using Exponential Gamma and Lindley Stress-Strength Distributions, Journal of Reliability and Statistics Studies, 5(1), p-33-41. Gogoi, J., Borah, M., and Sriwastav, G. L., 2010. An Interference Model with Number of Stresses a Poisson Process, IAPQR Transactions, 34(2), p139-152. Khan, Adil H., and. Jan, T. R., 2014a. Estimation of Multi Component Systems Reliability in Stress-Strength Models, Journal of Modern Applied Statistical Methods. 13(2), Article 21. Khan, Adil H., and. Jan, T. R., 2014b. Reliability Estimates of Generalized Poisson Distribution and Generalized Geometric Series Distribution, Journal of Modern Applied Statistical Methods. 13(2), Article 20. Khan, Adil H., and. Jan, T. R., 2014c. On estimation of reliability function of Consul and Geeta distributions, International Journal of Advanced Scientific and Technical Research. 4(4), p-96-105. Kotz, S., Lumelskii, Y. and Pensky, M., 2003. The Stress-Strength Model and its Generalizations: Theory and Applications, World Scientific Publishing, Singapore. Krishnamoorthy, K., Mukherjee, Shubhabarata and Guo, Huizhen, 2007. Inference on Reliability in two parameter exponential stress-strength model, Metrica, 65(3), p-261-273. Kundu, D. and Gupta, R.D., 2005. Estimation of P[Y < X] for generalized exponential distribution, Metrika, 61, p-291–308. Kundu, D. and Gupta, R.D., 2006. Estimation of R = P[Y < X] for Weibull distributions, IEEE Transactions on Reliability, 55, p-270–280. Raqab, M. Z. and Kundu, D., 2005. Comparison of Different Estimators of P[Y < X] for a Scaled Burr Type X Distribution, Communications in Statistics-Simulation and Computation, 34, p-465–483. Maiti, S. S., 1995. Estimation of Pr{X ≤ Y } in the Geometric case, Jour. Indian Statist. Assoc., 33, p- 87-91. Sandhya, K. and Umamaheswari, T. S., 2013. Estimation of Stress- Strength Reliability model using finite mixture of exponential distributions, International Journal of Computational Engineering Research, 3(11), p-39-46. Woodward, W. A.and Kelley, G. D., 1977. Minimum variance unbiased estimation of P(Y < X) in the normal case, Technometrics, 19, p-95-98.

[125]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Design of XOR Gate using Floating-Gate MOSFET Roshani Gupta*, Rockey Gupta, Susheel Sharma Department of Physics & Electronics, University of Jammu, Jammu, India

Abstract This paper presents the design of XOR gate using floating-gate MOSFET that is widely used technique for the design of mixed-mode circuits due to its unique feature of threshold voltage tunability through a bias voltage. The performance of CMOS XOR gate has been compared with FGMOS based XOR gate. It has been observed that by varying the bias voltages in FGMOS, the transient characteristics of XOR gate can be altered that results in less propagation delay and energy delay product as compared to CMOS XOR gate. The performance of these circuits has been verified through PSpice simulations carried out using level 7 parameters in 0.13 µm CMOS technology with a supply voltage of 1V.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Floating Gate MOSFET; CMOS; XOR Gate; Transient Characteristics; Propagation Delay; Energy Delay Product

1. Introduction The challenge of designing high performance low voltage and low power digital circuits is increasing due to the scaling down of CMOS technology and the increasing demand for portable electronic equipments. The general trend in CMOS technology is to make the devices smaller and smaller to increase the density and speed of digital circuits. It is also common to reduce the thickness of the gate oxide in order to increase the driving capability of the transistor. In addition, the thickness reduction implies that the supply voltage must be decreased to avoid excessive electric field in the devices as indicated by Chandrakasan et al., 1992 or Gonzalez et al., 1997 or Fayomi et al., 2004. The speed of conventional digital integrated circuits is degrading on reducing the supply voltage for a given technology. To fulfill these requirements, there is a need of development of new integrated circuits that have low voltage supply requirement, without any degradation in the performance. Yan et al., 2000 in his paper proposed various techniques for implementing low voltage and low power applications like Sub-threshold operation of MOSFET, Bulk-driven MOSFET, Level shifter technique, Self-cascode MOSFET, Floating-gate MOSFET and Quasi-floating-gate MOSFET. The Floating-gate MOSFET (FGMOS) proves to be a suitable element for low power applications as it allows programmability of threshold voltage and results in lowering of the threshold voltage below its conventional value. In this paper, we have employed floating-gate MOSFET (FGMOS) to implement XOR gate and compared its performance with its CMOS counter part. The performance of these circuits has been verified through PSpice simulations carried out using level 7 parameters in 0.13 µm CMOS technology with a supply voltage of 1 V. 2. Floating-Gate MOS Transistor Floating-Gate MOS transistor (FGMOS) is a modified form of conventional MOSFET where extra capacitances are introduced between the conventional gate and the multi-input signal gates as shown in Fig. 1. By applying a bias voltage on one of the input gates, the threshold voltage of FGMOS can be changed. FGMOS can be fabricated using a standard CMOS process by electrically isolating the gate of a standard MOSFET, so that there are no resistive

* Corresponding author. Tel.: +91 9796 264186. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Gupta et al/ COMMUNE– 2015

connections to its gate. A number of secondary gates or inputs are then deposited above the floating-gate (FG) which are electrically isolated from it and are only capacitively connected to FG. Since FG is completely surrounded by highly resistive material, so for DC operation, FG acts as floating node according to (Villegas, 2006), (Gupta et al., 2010), (Hasler et al., 2001), (Anand et al., 2013), (Keles et al., 2009) or (Murthy et al., 2011). VD CGD

V1

CGB

C1 FG

V2

VB

C2 CGS

VN CN

VS

Fig. 1 Floating-gate MOSFET

3. XOR Gate The XOR circuit is basic building block in various circuits, especially arithmetic circuits such as adders and multipliers, compressors, comparators, parity checkers, code converters, error-detecting or error-correcting codes and phase detectors according to (Keles et al., 2010) or (Wairya et al., 2012). The XOR (exclusive-OR) gate acts in the same way as the logical "either/or". In XOR gate, the output is high only if either, but not both of the inputs are one and the output is low if both inputs are one or if both inputs are zero i.e. the output becomes high if the inputs are different and low if the inputs are same. The circuit for CMOS XOR gate is shown in Fig. 2. VDD

A'

M1

A

M2

B

M3

B'

M4

Vout A

M5

A'

M6

B

M7

B'

M8

Fig. 2 XOR Gate

The performance of XOR gate can be characterized through its transient response which is a plot of input and output voltage with respect to time. Transient response is important to determine the maximum speed at which the device can be operated. It is measured between the 50% transition points of the input and output waveforms and is given by Hodges et al., 2005 or Razavi, 2008 as:

tP 

t plh  t phl

(1)

2

Where tplh defines the response time of the gate for a low to high output transition and tphl refers to the response time for a high to low output transition. The circuit of XOR gate has been simulated to obtain its transient characteristics by selecting W/L of p-channel MOSFETs as 2.6 μm/0.13 μm and n-channel MOSFETs as 1.3 μm/0.13 μm with the supply voltage of 1 V as shown in Fig. 3. From the simulation results, the propagation delay of XOR gate is calculated to be 0.3 ns.

[127]

Gupta et al/ COMMUNE– 2015

Vout (Volts)

1.2 1 0.8 0.6 0.4 0.2 0 0

1

2

3

4

5

Time (ns) Fig. 3 Transient characteristics of XOR gate

4. XOR Gate using FGMOS For high-speed digital circuits, logic gates must introduce a minimum amount of delay when inputs change. In order to enhance the speed of XOR gate, it is important to minimize the propagation delay. The propagation delay of XOR gate can be reduced by implementing the circuit using floating-gate MOS transistors (FGMOS) as shown in Fig. 4. The circuit is similar to CMOS XOR gate except that extra capacitances are introduced between the conventional gate and the input signal gate. The bias voltages Vbp and Vbn provide tunability to the threshold voltages of p and n-channel FGMOS transistors respectively. VDD

A'

C1 M1

A

C2 C3

Vbp B

A

B

M2 C6 C7

M3 C4 C9

B'

A'

C10 C11

Vbn

C5

M5 M7

C12

M4 C8

C14 C15 B'

Vout

C13 M6 M8 C16

Fig. 4 XOR gate using FGMOS

The simulation of circuit shown in Fig. 4 has been performed to obtain its transient response at different values of Vbp and Vbn with supply voltage of 1 V. Fig. 5 and 6 shows how propagation delay varies with bias voltage. In Fig. 5, bias voltage of p-channel FGMOS transistors (Vbp) is varied from 0 V to 1 V, while keeping bias voltage of n-channel FGMOS transistors (Vbn) fixed at 1 V. Similarly in Fig. 6, Vbn is varied from 0 V to 1 V, while keeping Vbp fixed at 0 V and output voltage (Vout) is obtained with respect to time. 1.2

Vout (Volts)

1 Vbp=0V Vbp=0.2V Vbp=0.4V Vbp=0.6V Vbp=0.8V Vbp=1V

0.8 0.6 0.4 0.2 0 0

1

2

3

Time (ns) Fig. 5 Transient response of XOR gate using FGMOS at different Vbp

[128]

4

5

Gupta et al/ COMMUNE– 2015 1.2

Vout (Volts)

1 Vbn=0V

0.8

Vbn=0.2V

0.6

Vbn=0.4V

Vbn=0.6V

0.4

Vbn=0.8V 0.2

Vbn=1V

0

0

1

2

3

4

5

Time (ns) Fig. 6 Transient response of XOR gate using FGMOS at different Vbn

Now, from the transient responses shown in Figs. 5 and 6, we have calculated the propagation delay at different bias voltages. The variation of propagation delay as a function of bias voltage is shown in Fig. 7. 0.6

Delay (ns)

0.5

Vbn

Vbp

0.4 0.3 0.2 0.1 0 0

0.2

0.4

0.6

0.8

1

Bias voltage (Volts)

Fig. 7 Propagation delay at different values of Vbp and Vbn

From the above results, it has been found that as we go on increasing the bias voltage of p-channel FGMOS transistor from 0 V to 1 V propagation delay increases from 0.18 ns to 0.57 ns, where as increasing bias voltage of nchannel FGMOS transistor from 0 V to 1 V reduces propagation delay from 0.39 ns to 0.18 ns. Thus, the appropriate selection of bias voltages of n and p-channel FGMOS transistors decreases the propagation delay, thus enhancing the operating speed. Now, the comparative transient characteristics of CMOS and FGMOS XOR gate have been obtained by selecting same W/L of M1, M2, M3 and M4 as 2.6 μm/0.13 μm and M5, M6, M7 and M8 as 1.3 μm/0.13 μm while keeping bias voltages of p and n-channel FGMOS transistors fixed i.e. Vbp = 0 V and Vbn = 1 V with supply voltage of 1 V and are shown in Fig. 8. From the simulation results, it has been found that FGMOS based XOR gate has propagation delay of 0.2 ns which is less as compared to CMOS XOR gate (tp = 0.3 ns).

Vout (Volts)

1.2 1

0.8

CMOS XOR

0.6

FGMOS XOR

0.4

0.2 0 0

1

2

3

4

5

Time (ns)

Fig. 8 Comparative transient response of XOR gate using CMOS and FGMOS

Now, the values of propagation delay obtained from the transient responses of XOR gate using CMOS and FGMOS has been used to calculate the energy delay product (EDP) at different values of supply voltage shown in Fig. 9. Energy delay product represents the trade off between power and the speed. Thus, FGMOS based XOR gate will exhibit more speed and dissipate less power as compared to CMOS XOR gate.

[129]

EDP (Js)*10 -23

Gupta et al/ COMMUNE– 2015

1.4 1.2 1 0.8 0.6 0.4 0.2 0

CMOS XOR FGMOS XOR

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

1

VDD (Volts) Fig. 9 Comparative EDPs of XOR gate using CMOS and FGMOS

From the graph shown above, it has been found that energy delay product varies with supply voltage and XOR gate implemented using FGMOS is better since the energy delay product is lower than CMOS XOR gate. 5. Conclusions In this paper, we have discussed the transient characteristics of XOR gate using CMOS as well as FGMOS. The performance of CMOS XOR gate has been compared with FGMOS. We have observed that variation in bias voltages of p and n-channel FGMOS transistors results in less propagation delay and energy delay product as compared to CMOS XOR gate. The performance of these circuits has been verified through PSpice simulations carried out using level 7 parameters in 0.13 µm CMOS technology with a supply voltage of 1V. References Chandrakasan, A. P., Sheng, S., Brodersen, R. W., 1992. Low-Power CMOS Digital Design, IEEE JSSC 27, p. 473. Gonzalez, R., Gordon, B. M., Horowitz, M. A., 1997. Supply and Threshold Voltage Scaling for Low Power CMOS, IEEE Journal of Solid-State circuits 32, p. 1210. Fayomi, C. J. B., Sawan, M., Roberts, G. W., 2004. Reliable Circuit Techniques for Low Voltage Analog Design in Deep Sub micron Standard CMOS: A Tutorial, Analog Integr. Circuits Signal Proc. 39, p. 21. Yan, S., Sinencio, E. S., 2000. Low voltage analog circuit design techniques: A Tutorial, IEICE Trans. Fundamentals E00-A, p. 1. Villegas, E. R., 2006. Low power and Low voltage circuit design using FGMOS transistor, IET Circuits, Devices and Systems series 20. Gupta, M., Pandey, R., 2010. FGMOS based voltage-controlled resistor and its applications, Microelectronics Journal 41, p. 25. Hasler, P., Lande, T. S., 2001. Overview of floating-gate devices, circuits and systems, IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing 48, p. 1. Anand, A., Mandal, S. K., Dash, A., Patro, B. S., 2013. FGMOS based low-voltage low-power high output impedance regulated cascode current mirror, International Journal of VLSI design & Communication Systems 4, p. 39. Keles, S., Kuntman, H. H., 2009. Four Quadrant FGMOS Multiplier, Proceedings of ELECO’: The 6th International Conference on Electrical and Electronics Engineering 2, p. 45–48. Murthy, P. H. S. T., Chaitanya, K., Krishna, M. M., Rao, M., 2011. FTL based 4Stage CLA Adder Design with Floating Gates, International Journal of Computer Applications 17, p. 1. Keles, F., Yildirim, T., 2010. Low voltage low power neuron circuit design based on subthreshold FGMOS transistors and XOR implementation, 11th International Workshop on Symbolic and Numerical Methods Modeling and Applications to Circuit Design, p. 1. Wairya, S., Nagaria, R. K., and Tiwari, S., 2012. Comparative Performance Analysis of XOR XNOR Function based high-speed CMOS Full Adder Circuits for low voltage VLSI Design, International Journal of VLSI design & Communication Systems (VLSICS) 3, p. 221. Hodges, D. A., Jackson, H. G., Saleh, R. A., 2005. Analysis and Design of Digital Integrated circuits, MC Graw-Hill. Razavi, B., 2008. Fundamentals of Microelectronics, John Wiley and sons.

[130]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Cellular Automata: Evolution and Parallel Dimensions Shah Jahan Wania*, M. A. Peera, K. A. Khanb a

Department of Computer Sciences, University of Kashmir, Srinagar, India b Govt. Degree College, Beerwah, Budgam, India

Abstract Cellular Automata rules producing evolution type phenomena have been used for a wide range of applications. Various models have been designed and explored for different applications. Although the strength of its parallelism has been felt by various researchers but its exploration for applications will not minimize the hardware but also maximize the optimum strength of processors. Our present study was intended to identify the additive 2D Cellular Automata linear rules on the quality of pattern evolution and the periodic parallelism utilization. We have made an analysis of 2DCA linear game of life (GOL) rule in Neumann neighborhood pattern evolution and observed pattern multiplication in the process. The results achieved will not only minimize the required hardware for parallel channel creation but also expand the microcomputer processing horizon.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Cellular Automata; Boundary Conditions; Computer Simulation; Patterns Grnrration; Introduction

Introduction Von Neumann and Stanislaw Ulam introduced cellular lattice in late 1940s as a frame work for modeling complex structures capable of self-reproduction (Von, 1966). Cellular Automata is based on a concept of dividing space into a regular lattice structure of cells where each cell can take a set of ‘n’ possible values. The value of the cell change in discrete time steps by the application of rule R that depends on the neighborhood around the cell. The neighborhood can be along a line, in a plane or in space. Cellular Automata (CA) model is composed of a universe of cells in a state having neighborhood and local rule. With the advancement of time in discrete steps the cell changes its value in accordance to the state of its neighbors. Thus the rules of the system are local and uniform. There are one- dimensional, two-dimensional and three-dimensional CA models. In one-dimensional CA the cells are like a linear canvas and the values of the canvas cells change due to application of a local rule in discrete advancing time steps. In two-dimensional CA the cells form a canvas plane and the changes take place in two dimensions while as in three dimensional CA volumetric changes take place by the application of local rule with advancement of time. As image is two dimensional data matrix, here we use 2DCA model, where cells are arranged in a two dimensional canvas matrix having interaction with neighboring cells. The central space represents the target cell (cell under consideration) and all spaces around represent its eight nearest neighbors. The structure of the neighbors mostly discussed and applied include Von Neumann neighborhood and Moore neighborhood, are shown in figure 1. In Von Neumann neighborhood, four cells are positioned at the orthogonal positions of the target cell (a i,j) while as Moore neighborhood is extension of Neumann structure with additional four cells placed diagonally at the four corner positions. For simplicity Von Neumann neighborhood cells can be termed as orthogonal neighbors and the additional cells by Moore can be called as corner neighbors.

*

Corresponding author. Tel.: +91 9086 897920. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Wani et al/ COMMUNE-2015

ai,j-1 ai-1,j

Target Cell

ai,j

ai+1,j

ai,j+1 Von Neumann Neighborhood

ai-1,j-1 ai,j-1

ai+1,j-1

ai-1,j

ai+1,j

Target Cell ai,j

ai-1,j+1 ai,j+1

ai+1,j+1

Moore Neighborhood Fig. 1 The two dimensional Cellular Automata in general are represented by the equation (I) as given below: [ai,j] t+1 = R[ ai, j , ai, j+1 , ai+1, j , ai, j-1 , ai-1, j ] t

--(I)

For Additive Cellular Automata the implementation of the famous totalistic rule in Von Neumann and Moore neighbourhoods, the respective representative equations can be written as follows: [ai,j] t+1 = XOR[ ai, j , ai, j+1 , ai+1, j , ai, j-1 , ai-1, j ] t --(II) [ai,j] t+1 = XOR[ ai, j , ai-1, j-1 , . . . . . . , ai+1, j+1 ] t --(III) Since the exploring worksheet/canvas is practically limited, researchers (Norman, 1986; Wolfram, 1985; Jarkko, 2012; Choudhury, et al) have defined some boundary conditions to facilitate the protection of data overflow outside the edges of the worksheet. On the basis of the applied boundary conditions the cellular automata have been divided into three main categories, briefly defined as follows: 1.1. Null Boundary Cellular Automata (NBCA) Under null boundary conditions the extreme edge cells are having zero values. For 1DCA the extreme right cell and the extreme left cell are considered to be having a value of binary zero ’0’ 1.2. Periodic Boundary Cellular Automata (PBCA) Under periodic boundary conditions the canvas is considered to be folded so that the extreme cells are taken to be adjacent to each other. For 1DCA the extreme right cell is considered to be adjacent to extreme left cell and the extreme left cell is considered to be adjacent to extreme right cell. 1.3. Intermediate Boundary Cellular Automata (IBCA) Under intermediate boundary conditions the left neighbor of the leftmost cell is regarded as the second right neighbor and right neighbor of the rightmost cell is considered as the second left neighbor. The linear additive 2-dimensional cellular automata attracted a number of researchers who have applied the rules for various applications in industry and research. The most important among such applications is the VLSI design that uses

[132]

Wani et al/COMMUNE – 2015

the cellular automata under the periodic boundary conditions. We have earlier made use of rules under null boundary conditions for cryptography and graphical translations (Wani, et al, 2014; Fasel, et al, 2012). The status value of the target cell at time t+1 depends on its own status value and the status value of cells in the Moore neighbourhood at time t, where t is earlier time than t+1. A number of research studies have been carried out by Stephen Wolfram and Norman H. Packard who categorized these rules under the discussed neighbourhoods into general, symmetric and totalistic. The famous example of 9-neighborhood totalistic cellular automata is John Horton Conway’s ‘Game of Life’. Use of the Game of Life rule in Von Neumann neighborhood has also been reported for pattern generation on multiple geometrical shapes by (Wani, et al, 2013) Various studies have also been carried out by Pabitra Pal Choudhury et al., who classified the cellular automata rules in Moore neighbourhood by assigning the rule values to different cells as shown in figure 2. The rules are generated by the interaction of target cell with itself and with the 8-neighbors around it. These nine rules are said to be basic or fundamental rules and group rules are derived from their combination, Group 2 are rules generated by addition of two basic rules, Group 3 by the combination of three basic rules, Group 4 by the combination of four basic rules, Group 5 by the combination of five basic rules and so on. Group 9 rule (only rule in the group) is the combination of all basic rules. All the combinations are additive (i.e. EX-OR operation).

64 128 32 1

256 2

Target Cell

16

8

4

Fig. 2: Examples: (Linear CA)

Rule 3 = Rule 2  Rule 1 Rule 11= Rule 8  Rule 2  Rule 1 Rule 15 = Rule 8  Rule 4  Rule 2  Rule 1

(Group 2) (Group 3) (Group 4)

Using this sort of configuration of patterns in 1DCA (Makoto, Leon, 2009; Radu, et al, 2006), we have reported (Wani, et al, 2013) 2DCA for cryptographic applications where the plaintext is converted to pattern based cipher. The cipher on the receiving end can be converted back to plaintext using the same CA rule in the forward iterations. This technique of generating ciphers has advantages of high cracking immunity due to wide range of rule possibilities and low hardware cost of implementation using VLSI technology. According to criteria of applying cellular automata rules to a group of data in any neighborhood, the cellular automata have been divided into two types: i) Uniform Cellular Automata ii) Non-Uniform Cellular Automata Uniform CA also known as Linear CA is where rule is applied uniformly on a data matrix of cells. All the cells in matrix get operated with the same rule. Non-Uniform CA also known as Hybrid CA is one in which all the cells of the matrix have their own local rule that may be different from the rule applied to other cells of that matrix. 2. Methodology For the experimentation purpose, we have used a null matrix of (129×129) elements and binary image of the seed was loaded at its centre. As the CA totalistic rule was applied it started generating patterns. The iteration loop control and the boundary conditions were also applied according to the following algorithm:

Label

Start Load a Null Matrix Input Binary Seed Start Iteration Counter Apply CA Rule Decrement Iteration Counter Loop to Label on Counter Condition Output Resulting Pattern Stop

[133]

Wani et al/ COMMUNE-2015

Start Load Seed Control Counter Boundary Condition Apply CA Rule No

Counter = 0 Yes Pattern Out Stop

Flowchart 3. Pattern Generation We have studied the pattern generation using totalistic cellular automata Game of Life rule (Rule 170) in Neumann neighborhood for the different orientations and have achieved results that could lead to various field applications for interdisciplinary research. In this new exploration we have used various geometrical orientations of the input seed and boundary conditions for the generation of the patterns in evolution as well as what can be termed as boundary invasion. The different orientations used range from simple regular shapes to boundary line matrices and deliberately introduced boundary defects. Different seed shape exploration generated the following pattern results on the MATLAB simulations under null boundary conditions and boundary invading conditions. 3.1. Pattern Generation Using Boundary Defects In this study, we have made the use of above methodology to observe the effect of null boundary on the world of full life i.e. universal matrix of ones. The dark or null boundary is a rectangle of zeros. Some resulting pattern generation references are indicated in Table 1 below: Table 1. Boundary Defect Patterns

T=0 (Dark World)

T=1 (Null Boundary)

T=10

T=21

T=29

T=37

T=41

T=47

T=53

[134]

Wani et al/COMMUNE – 2015

3.2. Pattern Generation Using Single Seed Boundary Invasion In this study same methodology is used to observe the effect of null boundary with a single central live seed on the world of full life i.e. universal matrix of ones. The dark or null boundary is a square of zeros with every side having a central seed of one. Some resulting pattern generation references are indicated in Table 2 below: Table 2. Single Seed Boundary Defect Patterns

T=0 (Dark World)

T=1 (Null Boundary)

T=10

T=21

T=29

T=37

T=41

T=51

T=59

3.3. Pattern Generation Using Single Seed Evolution and Boundary Invasion Moving ahead with the above methodology here we observe the effect of null boundary with a single central live seed at the centre of a world of full life i.e. universal matrix of ones. The dark or null boundary is a square of zeros. A single live seed is introduced at the centre of the matrix. Some resulting pattern generation references are indicated in Table 3 below:

[135]

Wani et al/ COMMUNE-2015 Table 3. Single Seed Evolution and Boundary Invasion Patterns

T=0 (Dark World)

T=1 (Null Boundary)

T=10

T=21

T=29

T=37

T=41

T=51

T=59

3.4. Pattern Generation Using Star Shaped Data Block Here we introduce a multi seed data block with the above methodology to observe the effect of evolution with the application of same rule. The evolution results in a periodic multiplication of the data block in various orientations. Some resulting pattern generation references are indicated in Table 4 below: Table 4. Multi Seed Evolution Patterns

T=0

T=16

T=24

T=32

[136]

Wani et al/COMMUNE – 2015

T=47

T=48

4. Discussion and Conclusion The results are definitely not only showing a bidirectional interaction that can be useful in studying physical processes under environmental conditions but also a powerful parallelism. The power of the parallelism can be utilized to minimize the hardware complexities.

References Choudhury, P. P., et al, Dec. 2010. Classification of CA Rules based on their Properties., IJCC. Fasel Qadir, Ahmad, P. Z., Wani, S. J., Peer M. A., December-2013. Quantum-Dot Cellular Automata: Theory and Applications, IEEE Conference on Machine Intelligence and Research Advancement (ICMIRA-2013), pp. 540-544. Fasel Qadir, Shah, J., Peer, M. A., and Khan, K. A., July 2012. Replacement of Graphic Translations with Two-Dimensional Cellular Automata, Twenty Five Neighborhood Model” IJCEM , pp. 33-39. Jarkko Kari, April, 2012. Universal Pattern Generation with Cellular Automata, Theoretical Computer Science, vol. 429, Makoto Itoh and Leon, O., 2009. Chua, Difference Equations for Cellular Automata, International Journal of Bifurcation and Chaos, Vol. 19 No.3, 805-830. Norman, H. Packard and S, 1986. Wolfram. Two Dimensional cellular Automata: J. Stat. Phys, vol. 38. Radu V., Craiu and Thomas, C., Lee, M., July 2006. Pattern Geeration Using Likelihood Interference for Cellular Automata IEEE Transactions on Image Processing vol. 15 No. 7. Shah, J. W., Fasel Qadir , Khan, K. A., and Peer, M. A., 2013, Dec. Springer Conference ICICIC Global Pattern Generation Using 2D Cellular Automata on Multiplr Geometrical Shapes. Von Neumann, J, 1966. Theory of Self-Reproducing Automata: University of Illinois Press. Wani, S. J., Khan, K. A., and Peer, M. A., April, 2014. 2D- Cellular Automata Linear Rules for Cryptography Based on Pattern Evolution. International Journal of Advanced Research in Computer Science Engineering and Information Technology vol 2 no.3. Wolfram, S., , 1985. Twenty Problems in the Theory of CA, Physica Scripta, pp. 170-183, vol T9.

[137]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16-18March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

High Impedance First-Order Transadmittance-Mode Allpass Filter Using CCII and OTA Nusrat Parveena*, Syed Zaffer Iqbalb a

Department of Electronics, Islamia college of Science and Commerece, Srinagar, India b Department of Physics,Government Women’s College, Nawakadak Srinagar, India

Abstract A novel transadmittance-mode (TAM) first-order allpass (AP) filter using a single second-generation current conveyor (CCII+), operational transconductance amplifier (OTA), one grounded resistor, and capacitor is presented. The input is voltage signal and output is current signal. The circuit has advantage of having input and output impedances high, thus facilitate cascading without additional devices. The phase angle, in addition to frequency of the applied signal, can also be adjusted electronically through the bias current of OTA without disturbing realizability condition. PSPICE simulation confirms the theoretical results

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Transadmittance Filter; High Impedance; Current Conveyor; Operational Transconductance Amplifier

1. Introduction Allpass filter is most important building block, used in analog signal processing. Due to potential advantages, these filters have received much attention among the circuit designers. AP filters are generally employed for introducing frequency dependent phase shift while keeping the amplitude of the input signal constant over the desired frequency range. Other areas of application of these active circuits are realization of oscillators and high-Q filters (T. Comer et al, 1997, Atti. Abuelma et al-1998, R. K. Rajeev et al, 2014, Toker . A. et al, 2000, A. Soliman, 1997). The active devices that have been used for the realizations of first-order all-pass circuits include operational amplifiers (OP-AMP), secondgeneration current conveyor (CCII), current feedback Op-Amps (CFOA), operational transconductance amplifier (OTA) and four-terminal floating nullor (FTFN). Several voltage and current-mode first-order all-pass filters are available in the literature (Higashimura M, 1990, Fukui Y et al,, 1999, O. Cicekoglu ety al, 2000, Khan, I. A and S. Maheshwari, 2000, Cam U.,et al, 2000, Khan,I. A and S Maheshwari, 2001, Pandey Paul. N, S.2006, Horng, J. et al, 2004, A Toker and S. Özoguz et al, 2001). A survey of technical literature reveals that there are only a few transadmittance-modes and/or Transimpdance-mode (TIM) filters (Toker A et al, 2001, Abuelma’atti M.T, Minaei, 2004, Shah, N. A et al 2004, Shah N. A et al, 2005, U Cam et al, 2005, Cam U, 2004). On the other hand, such type of filters are useful as an interface connecting a voltage-mode circuit to a current-mode circuit. Traditionally the current/voltage signal is converted to voltage/current signal and then processed. However, a TIM/TAM filter will perform this conversion and filtering simultaneously lending reduction in hardware, which in turn save chip area, thus resulting economically viable topologies. An application of transadmittance-mode filters for the realization of modern base-band receiver block of radio system can be found in (Toker A et al, 2001). Two first-order allpass filters one is TIM and other in TAM is reported in literature (Cam U et al, 2004, Cam U, 2005). The TIM employs one operational transresistance amplifier with the floating passive components and TAM uses one dual output current conveyor third generation (CCIII), two floating resistors, one grounded resistors and a grounded capacitor. The main purpose of this paper is to introduce a new transadmittance-mode first-order all-pass filter employing a single CCII+, one OTA, and each of grounded resistor and capacitor. Although the proposed circuit employs one more active component than the above mentioned first-order allpass filter, the proposed circuit provide the features of high input and output impedance *

Corresponding author. Tel.: +91 9419 426556. E-mail address: [email protected]

ISBN: 978-93-82288-63-3

Parveen and Syed/COMMUNE – 2015

and using only two grounded passive components. Since the TAM is the interface circuit used between voltage-mode and current-mode circuit thus it is essential for the TAM that input as well as output impedance should be high. The phase angle can be electronically adjusted through the bias current of OTA without disturbing the realizability condition Rg1= 1. Moreover, the circuit employs a grounded capacitor and resistor, besides offering electronic tunability feature, which is beneficial for IC technology (Horng, J et al, 2004). PSPICE simulation confirms the workability of the filter. 2.

The Proposed Circuit

Symbolic representation of second-generation current conveyor is shown in Fig 1(a), and is a all-round active device with versatile port relations given in the matrix form as under:  I y  0 0 0 V y  V   1 0 0  I   x   x   I x  0 1 0 V z 

(1)

OTA, circuit symbol is given in Fig. 1(b) and its micro modal in Fig. 1(c) is differential voltage controlled current source (DVCCS) and is characterized by the following port relation

(V   V  )  I o gm

(2)

where gm is the transconductance gain of OTA, used to electronically control the phase angle, is given by

gm 

I bias 2VT

where Ibias is the bias current and VT is the thermal voltage. Routine analysis of the proposed TAM filter shown in Fig. 1 yields the following transfer function:

Io 1  Vin g1

g2 ) g1 RC g (s  2 ) C

(s 

(3)

The realizability condition is g1R1 =1. The pole frequency is given by

o 

g2 C

(4)

and phase angle is given by

 (s)   2 tan 1 C / g 2 

(5)

Eq.(2) shows that the circuit realizes TAM first-order allpass transfer function. Moreover, from Eq.(4) phase can be controlled by g2 in addition to C and frequency of applied signal without disturbing realizability condition.

2. CCII Non-Idealities Taking the tracking errors of the CCII into account, the CCII is characterized by the following matrix

0 0 V y  I y   0 V     s  0 0  I x   x   I x   0   s  0 V z 

(6)

Where (s) and (s) represent the frequency transfers of the internal current and voltage followers of the CCII, respectively. They can be approximated by the following first-order functions (A Fabre et al, 1993)

[139]

Parveen and Syed/COMMUNE – 2015

 s  

And

 s  

0 1 s

0

0 1 s

0

where 0 = 0.9914,  = 3.8  109 rad/sec 0 = 0.9999,  = 6.48  109 rad/sec. If the proposed circuit is designed for the frequency much less than the corner frequency then, (s) and (s) take the form (s) = 1- i and (i Vth), the mobile charge increases with the increase in gate-source voltage. The results further show that the mobile charge increases due to the reduction in drain- source voltage.

[396]

Das et al/COMMUNE-2015

3.5

Comparison of n-CNTFET with 15nm Single Gate n-MOSFET and 60nm Trigate n-MOSFET

The comparison of ballistic n-CNTFET with 15nm n-MOSFET and 60nm Trigate n-MOSFET as shown in Table 1. It has been observed in table 1 that the ballistic n-CNTFET has very high drive current (Ion) which is 7295.55 % maximum than single gate MOSFET and 2819.29 % maximum than tri gate MOSFET. The ballistic n-CNTFET has low leakage current (Ioff) which is 79.44% lower than single gate MOSFET and 47.14 % lower than tri gate MOSFET. The ballistic n-CNTFET has high Ion/Ioff ratio (~106) i.e. fast switching speed which is maximum as compared to 15nm n-MOSFET (103) and 60nm Trigate n-MOSFET (104). The short-channel effects such as Subthreshold Slope (SS) and Drain Induced Barrier Lowering (DIBL) are ideal in case of ballistic n-CNTFET. Hence, we can say that the controllability of gate over the channel will be more in case of ballistic n-CNTFET. It is further observed that the CNTFET is an excellent way for further extension of Moore’s Law. Table 1. Comparison of ballistic CNTFET with single gate MOSFET and tri gate MOSFET.

Ion

15nm MOSFET (Chau et al., 2003) 450 µA/µm

60nm Tri gate MOSFET (Chau et al., 2003) 1140 µA/µm

Ioff

180 nA/µm

70 nA/µm

37 nA/µm

Ion/Ioff

2500

16285.71

898730

SS

95 mV/decade

68 mV/decade

60 mV/decade

DIBL

100 mV/V

41 mV/V

0.60 mV/V

Device Metrics

Ballistic CNTFET with CNT diameter = 1nm 33, 280 µA/µm

4. Conclusion This paper investigates the performance of cylindrical shaped ballistic n-CNTFET. It can be concluded that the CNTFET perform better when the gate or drain voltage is increased and it has very high drive current (Ion), low leakage current (Ioff), high Ion/Ioff ratio (~106) i.e. fast switching speed and improves short channel effects in comparison with single-gate and tri gate MOSFET. 5. Acknowledgements One of the authors Mr. Devi Dass gratefully acknowledges the University Grants Commission (U.G.C.), for the award of ‘Rajiv Gandhi National Fellowship’ under the scheme funded by Ministry of Social Justice & Empowerment, Govt. of India. References Chau, R., Boyanov, B., Doyle, B., Doczy, M., Datta, S., Hareland, S., Jin, B., Kavalieros, J., Metz, M., 2003. Silicon Nano-transistors for Logic Applications, Physica E 19, p. 1. Lundstrom, M., 2003. Moore's Law Forever?, Science 299, p. 210. Theis, T. N., Solomon, P. M., 2010. It's Time to Reinvent the Transistor, Science 327, p. 1600. Hasegawa, H., Kasai, S., Sato, T., 2004. Hexagonal Binary Decision Diagram Quantum Circuit Approach for Ultra-Low Power III-V Quantum LSls, IEICE Transaction on Electron, E87-C, 1757 (2004). Hashim, A.M., Pung, H.H., Pin, C.Y., 2008. Characterization of MOSFET-like Carbon Nanotube Field Effect Transistor, Jurnal Teknologi 49 (D), p. 129. http://www.public.itrs.net Prasher, R., Dass, D., Vaid, R., 2013. Study of Novel Channel Materials Using III-V Compounds with Various Gate Dielectrics, International Journal on Organic Electronics (IJOE) 2, p. 11. Prasher, R., Dass, D., Vaid, R., 2013. Performance of a Double Gate Nanoscale MOSFET (DG-MOSFET) Based on Novel Channel Materials, Journal of Nano and Electronic Physics 5, p. 010171. Ferain, I., Colinge, C. A., Colinge, J.P., 2011. Multigate Transistors as the Future of Classical Metal-Oxide-Semiconductor Field-Effect Transistors, Nature 479, p. 310. Dass, D., Prasher, R., Vaid, R., 2013. Impact of Scaling Gate Insulator Thickness on the Performance of Carbon Nanotube Field Effect Transistors (CNTFETs), Journal of Nano and Electronic Physics, 5, p. 020141. Tans, S. J., Verschueren, A.R.M., Dekker, C.,1998. Room-Temperature Transistor Based on a Single Carbon Nanotube, Nature 393, p. 49. Martel, R., Schmidt, T., Shea, H. R., Hertel, T., Avouris, P., 1998. Single-and Multi-Wall Carbon Nanotube Field-Effect Transistors, Applied Physics Letters 73, p. 2447. Javey, A., Guo, J., Wang, Q., Lundstrom, M., Dai, H., 2003. Ballistic Carbon Nanotube Field-Effect Transistors, Nature 424, p. 654. Rahman, A., Wang, J., Guo, J., Hasan, M. S., Liu, Y., Matsudaira, A., Ahmed, S. S., Datta, S., Lundstrom, M., 2006. FETToy.

[397]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Morphological Analysis of Proper Nouns in Punjabi Umrinderpal Singh*, Vishal Goyal, Gurpreet Singh Lehal Department of Computer Science Punjabi University Patiala, India

Abstract The Morphological is the branch of linguistics. This field is related to the study of structure of words. The Morphological Analysis becomes a very popular branch of research in all languages, especially for the morphological rich languages like Indo-Aryan languages. The Morphological Analysis and Generator play essential roles in many Natural Language Processing (NLP) applications like Part of Speech (POS) Tagger, Named Entity Recognitions (NER) and many other applications. This paper presents an approach to Morphological Analysis of Proper Nouns in Punjabi. A thorough analysis has been done for Punjabi and found many suffix patterns that can be used to classify Proper Nouns. Two thousand thirty unique suffixes has been found during analysis. Analysis has been done on a large Punjabi Corpus having 278098 unique words. Based on these suffixes proposed algorithm was able to get 97.42% accuracy to identify proper nouns.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: NLP: Part of Speech Tagger; POS; Punjabi; Rule Based; HMM

1.

Introduction

The Morphological analysis and generation is an essential part of any NLP applications. It has a vital role in morphological rich languages such as Punjabi, Hindi and other Indo-Aryan languages. In morphological analysis, study of the structure of the words based on its root and affixes. The morphological analysis has become essential in IndoAryan languages, where words can be inflected in many forms and yield different meaning. The main concern of the morphological analysis to get the grammatical information by studying the word's structure like gender, number person etc.[Goyal and Lehal. 2008]. A word can be of two types; simple and compound. Where simple words combination of root words and its suffixes on the other hand compound words can be broken into two independent words and there independent words has their own meaning[ Goyal and Lehal 2008]. Inflection morphology [Bansal et.al 2011] gives us different form of a word by adding or removing affixes. Changes in the word meaning are minimal for exp: Cat Faster

---

Cats Fast

In derivational morphology, derives new forms of words from existing words and word class is changed on deriving for example: Modern (adj) Drink (v) --

-Modernize (v) Drinkable (adj)

2. Related Work Punjabi is one of the rich Morphological languages in Indo-Aryan language's family. However, very few morphological analyzer and generator has been developed for Punjabi. [Gill 2007] had been developed rule based morphological analyzer for Punjabi. This system was developed for Punjabi grammar checker. The system was based on database lookup for root words. If word not found in database, it marks it as unknown [Bansal et.al 2011]. [Singh *

Corresponding author. Tel. +91 9914996719 E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Singh et al/ COMMUNE-2015

and Kumar 2013] has been developed a Punjabi Morphological analyzer for Nouns, Pronouns and Adjectives. This Morphological analyzer was the part of UNL (Universal Network Language) based Machine Translation system (MT) system. [Bansal et.al 2011] suggests way to increase the accuracy of Morphological analyzer system developed by [Gill et.al 2007],[Mohd. Hunayoun et. Al] has been described as an implementation of morphological and development of corpus for Punjabi in Shamukhi script, which is spoken in Pakistan Punjab. [Goyal and Lehal 2008] had been developed morphological analyzer for Hindi as part of the MT system Hindi to Punjabi. [Snigdh paul et.al 2013] has been proposed a method for over stemming word in Hindi by various rules to make the word proper stem. Reader may fallow [Shweta Vikram 2013; Antony PJ et. al. 2012] paper for more details of morphological analysis for Indian languages and available techniques. Related word shows that very limited work has been done for Indian languages, especially for Punjabi. One should need to explore all morphological rich languages in detail. Efforts were made to explore Proper Nouns in Indian Languages and shows that how rich are these languages even for Proper Nouns. To best of our knowledge, we did not find any work that particular based on morphological analysis of proper nouns in Indo-Aryan languages. 3. Morphology of Proper Nouns in Indian Languages Indian languages have very rich morphological features as compared to European languages. Words can be inflected in many forms and may yield different meanings. Many researcher has been proposed various algorithm and methods only for open class words like nouns, verb and adjectives etc. [Rohit kansal 2012; Harinder Singh 2013; Mayuri Rostagi at.al 2014]. However, none of the researcher works on morphological analysis of Proper Nouns. By analysis of proper nouns in Indian Languages (IL) we can found various hidden features or information which can help us to classify the word as proper noun. The Morphological analysis of proper nouns is essential for resource poor languages. It is not possible to collect all the proper nouns for a particular language and proper nouns may came in any form. Most of the proper nouns are ambiguous for example word ਲਾਲ (Lal)(Red) is a valid noun word and it also valid person name. In this paper, morphological analysis has been done for proper nouns in Punjabi language and find out various suffixes, which were part of the person and location names in Punjabi. By doing the analysis of these suffix patterns, one can identify the word as proper noun. This kind of analysis can be very useful in POS tagging, where algorithm can check, out of vocabulary (OOV) words and classify them into proper nouns. 4 Patterns in Proper Nouns Morphological rich languages like Punjabi; Hindi Urdu has some unique features related to proper nouns. Analysis has been done to find out these unique suffixes in proper nouns. Most of the proper nouns in Indo-Aryan languages consist of two words, root word and its suffix for example. person name 'Ramdas' consist of two independent words 'Ram' and 'Das' where 'Ram' means 'lord Rama ' and 'das' means 'Servant' combined meaning of both words servant of load Rama. Based on these suffixes we can able to find out all proper names that ends with das. Following table shows proper nouns suffixes. Table 1. Compound word for Proper Nouns S.No. 1

Name ਜੀਵਨਦਾਸ(Jīvanadāsa)

Suffix ਦਾਸ(Dāsa)

2

ਹਰਨਾਮਦਾਸ(Haranāmadāsa)

ਦਾਸ(Dāsa)

3

ਹਰਰਦਾਸ(Haridāsa)

ਦਾਸ(Dāsa)

4

ਜੀਵਨਦਾਸ(Jīvanadāsa)

ਦਾਸ(Dāsa)

5

ਰਰਵਦਾਸ(Ravidāsa)

ਦਾਸ(Dāsa)

6

ਮੁਰਨਦਾਸ(Munidāsa)

ਦਾਸ(Dāsa)

Table 1 shows compound words consists of two independent words. There are many proper nouns, which have simple word forms, consist of root and their suffix but have unique features to identify them as proper noun. Following table 2 shows simple form of words, which are valid candidate words for proper nouns. Table 2.Simple words form for Proper Noun S.No. 1

Name ਤਾਰਰਕਾ(Tārikā)

Suffix ਾਾਰਰਕਾ(̔Ārikā)

2

ਸਾਰਰਕਾ(Sārikā)

ਾਾਰਰਕਾ(̔Ārikā)

3

ਹਜਾਰਰਕਾ(Hajārikā)

ਾਾਰਰਕਾ(̔Ārikā)

4

ਦਵਾਰਰਕਾ(Davārikā)

ਾਾਰਰਕਾ(̔Ārikā)

[399]

Singh et al/ COMMUNE-2015

S.No. 5

Name ਸਾਰਰਕਾ(Śārikā)

Suffix ਾਾਰਰਕਾ(̔Ārikā)

6

ਰਨਹਾਰਰਕਾ(Nihārikā)

ਾਾਰਰਕਾ(̔Ārikā)

By performing an analysis on Punjabi language and find out 230 categories, which were unique to Punjabi language and no evidence where found where these suffixes conflict with other words of Punjabi language. Most categories of suffixes have unique patterns but few categories were also found those suffixes might conflict with other Punjabi words, which were not person names for example; word ਉਦਾਸ has valid suffix ਦਾਸ but may or may not be person name. These words were ambiguous for names category. All these words treated as special cases and tried to resolve their ambiguity by checking previous and next word of current position. The person designation list like ਸਰੀ, ਸਰੀਮਾਨ, ਸਾਰਹਬ etc. and last name ਕੁਮਾਰ, ਰਸਿੰ ਘ etc. has collected to resolve ambiguity of these words. If ambiguous words start with some designation or end with some last name based on this localized information we can resolve their ambiguity. A proper name may have two valid suffixes; we have used longest suffixes stripping approach to identify the word. Following table shows some names having two suffixes. Table 3. Two or more valid suffixes for Person Names S.No. 1

Name ਅਜਮੁੁੱ ਦੀਨ(Ajamudīna)

Suffix1 ਾੁੱਦੀਨ

Suffix2 ਦੀਨ

2

ਅਨਵਰੁੁੱ ਦੀਨ(Anavarudīna)

ਾੁੱਦੀਨ

ਦੀਨ

3

ਉਮਾਵਤੀ(Umāvatī)

ਾਾਵਤੀ

ਵਤੀ

4

ਅਮਰਾਵਤੀ(Amarāvatī)

ਾਾਵਤੀ

ਵਤੀ

5

ਅੋਮਾਨਿੰਦ(A̔ōmānada)

ਾਾਨਿੰਦ

ਨਿੰਦ

6

ਆਤਮਾਨਿੰਦ(Ātamānada)

ਾਾਨਿੰਦ

ਨਿੰਦ

7.

ਕਮਲੇ ਸ਼ਵਰੀ(Kamalēśavarī)

ਾੇਸ਼ਵਰੀ

ਸ਼ਵਰੀ

8.

ਸੁਰੇਸ਼ਵਰੀ(Surēśavarī)

ਾੇਸ਼ਵਰੀ

ਸ਼ਵਰੀ

9.

ਰਾਮਪ੍ਰਸਾਦ(Rāmaprasāda)

ਪ੍ਰਸਾਦ

ਸਾਦ

10.

ਪ੍ੀਰਪ੍ਰਸਾਦ(Pīraprasāda)

ਪ੍ਰਸਾਦ

ਸਾਦ

Stripping the largest suffix in to identify the proper nouns always is better clue as compared to smallest suffix stripping. Smallest suffix striping may yield ambiguous information to identify proper nouns, for example. ਰਾਮਪ੍ਰਸਾਦ name has two valid suffixes ਪ੍ਰਸਾਦ and ਸਾਦ has length 5 and 3 respectable. Suffix ਪ੍ਰਸਾਦ always helps to identify word as proper name and there is no chance to classify word wrongly, but second suffix ਸਾਦ is also a valid name to identify word ਪ੍ਰਸਾਦ but has high chances to identify word wrongly for example, ਫ਼ਸਾਦ word also ends with suffix ਸਾਦ but ਫ਼ਸਾਦ is not a person name. Therefore, always better to check longest suffix first to identify proper nouns. Like person names, location names also have morphological features to identify them. Following table 4 shows location names that end up some particular suffixes. Table 4. Location names with suffixes S.No 1

Location Names ਸੁਲਤਾਨਪ੍ੁਰ(Sulatānapura)

Suffix ਪ੍ੁਰ(Pura)

2

ਹੈਦਰਾਬਾਦ(Haidarābāda)

ਾਾਬਾਦ(Ābāda)

3

ਰਕਸਨਪ੍ੁਰਾ(Kisanapurā)

ਪ੍ੁਰਾ(Purā)

4

ਉਜਬੇਰਕਸਤਾਨ(Ujabēkisatāna)

ਸਤਾਨ(Satāna)

5

ਮਨੀਮਾਜਰਾ(Manīmājarā)

ਮਾਜਰਾ(Mājarā)

5. Methodology used to Extract Proper Nouns In-depth analysis has been done on the Punjabi corpus to extract all valid suffixes for Proper Nouns, 80% of the total corpus has been used for this analysis. Following diagram 1 shows process to extract suffixes for proper noun. Initially we have collected around 30000 proper nouns related to person names and locations. All person names belong to different communities of Indo-Aryan cultures.

[400]

Singh et al/ COMMUNE-2015 Table 5.Training Corpus Detail Domain News Data Health Data Tourism Data Agriculture Data

Total Words 400200 200541 230853 208422

Unique Words 150040 60087 63082 72421

Along with this proper noun lexicon, a huge raw corpus of Punjabi was used to find out suffix patterns related to names. Table 5 shows the details of corpus. A huge raw corpus was needed to check proper noun suffixes, to extract only those suffix patterns, which were unique in all domains and words, should not conflict with any domain's data. We have developed the algorithm to extract all unique suffix patterns along with their frequency and suffix length. All unique words have been extracted from corpus of 1040016 words. We have found 278098 total unique words. All unique suffixes having length greater than 3 applied on these unique words, to extract all those words that end with given suffixes. We have chosen the length greater than 3 because length less than this yield many ambiguous words, which were not candidate words for proper nouns. After applying all suffixes, we got 47501 words, which were 17% of all the unique words. All proper nouns extracted from 47501 words, which already part of the proper names lexicon, along with their suffixes, and we left with two lists. The List having new words, which may be the candidates for proper nouns and list of words, which we already have in our proper name lexicon. We have divided the list of all words end with valid suffix into two lists because it made us easier to find out all ambiguous words. Proper Nouns

Raw Corpus

Suffixes length >=3 Extract Unique Suffixes

Prop er

Extract Unique Words

Find Valid

Words Ends with Suffixes

Valid Names with suffixes

No

New Names/ Words with suffixes

Yes

If Suffixes Common in both list

Ambiguous suffixes for Names

Valid suffixes for Names

Tag and remove all Verbs

List of Verbs

Manually check all names and suffixes

Valid list of suffixes for Proper Nouns

Fig. 1: Methodology to extract valid suffixes

Compassion of both list's suffixes yield all ambiguous words having common suffix in both lists. If suffix is only part of the proper name list, we can choose that suffix as a valid suffix to identify proper noun without any ambiguity. On the other hand, we had a list of ambiguous words and their suffixes. To refine ambiguous word list, we had used the list of verbs. The verb tag helps us to tag and remove all those words, which are verbs. The verb list consists of around

[401]

Singh et al/ COMMUNE-2015

23000 unique words. All forms of the verb have been used to tag verbs in list. Following table 6 shows the list of Verb tags used for tagging the verbs in ambiguous list. Table 6.All Verb POS tags Verb Types Main Verb Non-Finite Infinite Gerund Auxiliary

Label V_VM V_VM_VNF V_VM_VINF V_VM_VNG V_VAUX

Proper nouns mainly confessed with Adjectives and Nouns. Therefore, we need to check all these words manually. Whether they were valid words for proper nouns or not. Manually analysis has been done to extract all those suffixes, which can be used to classify words as proper nouns. Ambiguous word list also extracted that ends with valid suffixes. 6. System Architecture to Find Proper Nouns In out proposed system to find proper noun based on suffixes, four different lists were used to identify names correctly. Initially, the systems take Punjabi input text. The tokenization and normalization process token the text into words and remove unwanted symbols. The system find and remove all ambiguous words from the input text like word ਰਜਿੰ ਦਰਾ end with valid suffix ਾਿੰਦਰਾ but never used as person name. Suffix ਾਿੰਦਰਾ used to identify many other names like ਵਰਰਿੰ ਦਰਾ, ਰਰਜਿੰ ਦਰਾ, ਭੁਰਪ੍ਿੰ ਦਰਾ. Therefore, we have developed a list of words, which were mostly nouns to identify ambiguous words. This list contains 203 ambiguous words. Following table 7 shows some of the ambiguous words having valid suffixes. Table 7. Ambiguous words SNo. 1

Names ਮਨੋਰਿੰ ਜਨ(Manōrajana)

Suffix ਰਿੰ ਜਨ(Rajana)

2

ਹਸਤਪ੍ਾਲ(Hasatapāla)

ਤਪ੍ਾਲ(Tapāla)

3

ਪ੍ੁਸਰਤਕਾ(Pusatikā)

ਰਤਕਾ(Tikā)

4.

ਜੀਵਨ(Jīvana)

ਜੀਵਨ(Jīvana)

5.

ਏਕਾਂਤ(Ēkānta)

ਕਾਂਤ(Kānta)

The list of valid suffixes applied on all tokens to identify proper nouns and mark them for further processing. We have used 230 suffixes to identify proper names. Appendix A shows some of the most frequently used suffixes for proper nouns. Input Text

Tokenization/Normali zation

List of Ambiguous words which are not names

Find and Remove ambiguous words

Compare words with valid suffix list

List of Ambiguous names

Find and Tag ambiguous names

Rules to resolve ambiguity Fig 2: System Architecture Output Text

[402]

List of valid suffixes

List of Designation s middle names and last names

Singh et al/ COMMUNE-2015

The list of ambiguous proper nouns has been used to identify and mark them. Ambiguous names like word ਸਮੁਿੰ ਦਰ, ਜੀਵਨ etc. can be used as proper nouns or they can be used can adjective and noun. The algorithm marks them as ambiguous names and various rules have been used to resolve their ambiguities. The rules used localization information and list of designation, middle names and last names to resolve their ambiguities for example word ਸਮੁਿੰ ਦਰ is an ambiguous word, which can be used as proper names or can be used as adjective. ਫੁੁੱ ਲ ਬਹੁਤ ਹੀ ਸੁਿੰ ਦਰ ਹੈ.( Flower is very beautiful) ਸਰੀ ਸੁਿੰ ਦਰ ਕੁਮਾਰ (Shree Sunder Kumar) Algorithm check previous and next words to disambiguate it like if it has some designation ਸਰੀ or ਸਰੀ ਮਾਨ etc or may have surname ਕੁਮਾਰ, ਰਸਿੰ ਘ etc. then it marked as proper noun by algorithm. The system output is all valid proper names based on their suffix information. 7 Evaluation valuation of the all the suffixes has been done on 20% of the corpus which was not used for suffix pattern extraction. Testing data consists of four domains namely political news data, Heath data, Tourism data and agriculture data. Table 8 shows the detail of testing data. Table 8. Testing Data Details Domain News Data Health Data Tourism Data Agriculture Data

Total Words 30300 8998 5859 9893

Total unique words in test data was 2502. Testing data contains 453 words ends with valid suffixes used for proper noun classification. Out of 453 words, 54 words were ambiguous. These words ends with valid suffix but words can be classified as noun, adjectives or proper nouns. Evaluation has been done using standard metrics. Recall(R) =

𝐂𝐨𝐫𝐫𝐞𝐜𝐭 𝐚𝐧𝐬𝐰𝐞𝐫𝐬 𝐠𝐢𝐯𝐞𝐧 𝐛𝐲 𝐬𝐲𝐬𝐭𝐞𝐦 𝐓𝐨𝐭𝐚𝐥 𝐩𝐨𝐬𝐬𝐢𝐛𝐥𝐞 𝐜𝐨𝐫𝐫𝐞𝐜𝐭 𝐚𝐧𝐬𝐰𝐞𝐫𝐬

Precision (P) = F1-Measure =

𝐂𝐨𝐫𝐫𝐞𝐜𝐭 𝐚𝐧𝐬𝐰𝐞𝐫𝐬 𝐀𝐧𝐬𝐰𝐞𝐫𝐬 𝐩𝐫𝐨𝐝𝐮𝐜𝐞𝐝

𝟐∗𝐑∗𝐏 𝐑+𝐏

Table 9. System evaluation of system

Recall Precision F1-Measure

97.13 97.72 97.42

Algorithm yield 97.42% accuracy to find proper nouns based on their suffixes. Sometime the system failed to find some proper nouns because of their ambiguities. Some person names were ambiguous and have no information around them in sentence to resolve their ambiguity like designation and surnames. The system also marks some words, which were not person names like word ਅਰਨਿੰਦਰਾ having valid suffix ਾਿੰਦਰਾ. 8 Conclusion The morphological analyzer has been used by many researchers in different NLP applications like POS tagger and NER systems. It’s role becomes very essential in morphological rich languages. In this paper, an approach has been discussed to find proper based on their suffix information. Analysis shows that how rich Indo-Aryan languages are for

[403]

Singh et al/ COMMUNE-2015

even proper nouns. To the best of our knowledge, there has been no work done for proper nouns, analysis based on their morphological information. Various morphological analyzers and generators have been proposed by the researcher for Indo-Aryan languages. These analyzers were based on verbs, nouns and other part of speech. Algorithm successfully extracted many suffix patterns for Punjabi and used them to extract proper nouns with the accuracy of 97.42%. Most of the suffixes can be directly used in other Indian languages without any conflicts with other words of that language. In future work, effort will made to explore Hindi and Urdu languages for Proper Name's suffixes. References Antony P J and Dr K P Soman. 2012. Computational Morphology and Natural Language Parsing for Indian Languages: A Literature Survey, Volume 3 No. 4: 136-146 Deepak Kumar Manjeet Singh and Seema Shukla. 2012. FST Based Morphological Analyzer for Hindi, volume 9, 349-353 Gagan Bansal, Satinder Pal Ahuja, Sanjeev Kumar Sharma. 2011. Improving Existing Punjabi Morphological Analyzer, Volume 5: 221-229 Gill Mandeep Singh, Lehal Gurpreet Singh, Joshi S.S. 2007. A full form lexicon based Morphological Analysis and generation tool for Punjabi, International Journal of Cybernatics and Informatics, Hyderabad, India,October 2007, pp 38-47 Harinder Singh and Parteek Kumar. 2013. Analysis of Noun, Pronoun and Adjective Morphology for NLization of Punjabi with EUGENE, volume 2: 436-442 Mohd. Shahid Husain. 2012. An Unsupervised Approach to Develop Stemmer, International Journal on Natural Language Computing, 15-23 Manish Shrivastava and Pushpak Bhattacharyya. 2008. Hindi POS Tagger Using Naive Stemming : Harnessing Morphological Information Without Extensive Linguistic Knowledg, International Conference on NLP (ICON08), Pune, India, December, 2008 Also accessible from http://ltrc.iiit.ac.in/proceedings/ICON-2008 Mayuri Rastogi and Pooja Khanna. 2014. Development of Morphological Analyzer for Hindi, International Journal of Computer Applications Volume 95-No-17: 1-5 Muhammad Humayoun and Aarne Ranta. 2010. Developing Punjabi Morphology, Corpus and Lexicon, n. The 24th Pacific Asia conference on Language, Information and Computation, : 163-172 Snigdha Paul, Mini Tandon, Nisheeth Joshi And Iti Mathur. 2013. Design Of A Rule Based Hindi Lemmatizer: 67-74 Vaishali Gupta, Nisheeth Joshi and Iti Mathur. 2013. Rule Based Stemmer in Urdu, Computer and Communication Technology (ICCCT), 129 - 132 Vishal Goyal, Gurpreet Singh Lehal. 2008. Hindi Morphological Analyzer and Generator, First International Conference on Emerging Trends in Engineering and Technology: 1156-1159. Appendix A. List of frequently used suffixes S.No. 1.

Suffixes

Examples

ਨਪ੍ਰੀਤ(Naprīta)

ਹਰਮਨਪ੍ਰੀਤ,ਮਨਪ੍ਰੀਤ(Haramanaprīta,manaprīta)

2.

ਨਾਰਾਇਣ(nārā'iṇa)

ਜੈਨਾਰਾਇਣ,ਰਸਵਨਾਰਾਇਣ(ainārā'iṇa,śivanārā'iṇa)

3.

ਪ੍ਰਸਾਦ(prasāda)

ਰਸ਼ਵਪ੍ਰਸਾਦ, ਰਾਮਪ੍ਰਸਾਦ(śivaprasāda, rāmaprasāda)

4.

ਪ੍ਰਕਾਸ਼(prakāśa)

ਜੈਪ੍ਰਕਾਸ਼, ਓਮਪ੍ਰਕਾਸ਼(jaiprakāśa, ōmaprakāśa)

5.

ਰਕਸ਼ੋਰ(kiśōra)

ਨਿੰਦਰਕਸ਼ੋਰ,ਰਾਜਰਕਸ਼ੋਰ(nadakiśōra,rājakiśōra)

6.

ਕੁਮਾਰ(kumāra)

ਰਸ਼ਵਕੁਮਾਰ,ਰਾਜਕੁਮਾਰ(śivakumāra,rājakumāra)

7.

ਗੋਪ੍ਾਲ(gōpāla)

ਰਕਰਸ਼ਣਗੋਪ੍ਾਲ,ਰਾਮਗੋਪ੍ਾਲ(kriśaṇagōpāla,rāmagōpāla)

8.

ਚਿੰ ਦਰ(cadra)

ਰਾਮਚਿੰ ਦਰ,ਖੇਮਚਿੰ ਦਰ(rāmacadra,khēmacadra)

9.

ਰਜਿੰ ਦਰ(jidara)

ਰਾਰਜਿੰ ਦਰ,ਰਰਜਿੰ ਦਰ(rājidara,rajidara)

10.

ਜੇਂਦਰ(jēndara)

ਰਾਜੇਂਦਰ,ਬਰਜੇਂਦਰ(rājēndara,brajēndara)

11.

ਾਿੰਰਤਕਾ(tikā)

ਅਿੰ ਰਤਕਾ,ਅਵਿੰ ਰਤਕਾ(atikā,avatikā)

12.

ਪ੍ਰੀਤ(prīta)

ਰਦਲਪ੍ਰੀਤ,ਹਰਪ੍ਰੀਤ(dilaprīta,haraprīta)

13.

ਪ੍ਾਲ(pāla)

ਸੁਖਪ੍ਾਲ,ਸੁਰਰਿੰ ਦਰਪ੍ਾਲ(sukhapāla,suridarapāla)

14.

ਰਪ੍ਿੰ ਦਰ(pidara)

ਨਰਪ੍ਿੰ ਦਰ,ਹਰਰਪ੍ਿੰ ਦਰ(napidara,harapidara)

15.

ਪ੍ੁਰ(pura)

ਸਰਹਬਾਜ਼ਪ੍ੁਰ,ਫੁੱ ਤੇਪ੍ੁਰ(sahibāzapura,phatēpura)

20.

ਮੀਤ(mīta)

ਗੁਰਮੀਤ,ਪ੍ਰਮੀਤ(guramīta,paramīta)

21.

ਮੇਸ਼(mēśa)

ਰਮੇਸ਼,ਸੋਮੇਸ਼(ramēśa,sōmēśa)

22.

ਮੇਂਦਰ(mēndara)

ਧਰਮੇਂਦਰ,ਉਮੇਂਦਰ(dharamēndara,umēndara)

23.

ਮੋਹਨ(mōhana)

ਮਨਮੋਹਨ,ਰਾਜਮੋਹਨ(manamōhana,rājamōhana)

24.

ਰਿੰ ਜਨ(rajana)

ਨਰਿੰ ਜਨ,ਰਸਵਰਿੰ ਜਨ(narajana,śivarajana)

25.

ਰਜੀਤ(rajīta)

ਸੂਰਜੀਤ,ਇਿੰ ਦਰਜੀਤ(sūrajīta,idarajīta)

26.

ਾੇਾਦਰ(̔ēndara)

ਜੀਤੇਂਦਰ,ਸ਼ੈਲੇਂਦਰ(jītēndara,śailēndara)

27.

ਜੋਤ(jōta)

ਮਨਜੋਤ,ਨਵਜੋਤ(manajōta,navajōta)

28.

ਤਪ੍ਾਲ(tapāla)

ਪ੍ਰਨੀਤਪ੍ਾਲ,ਸੁਖਜੀਤਪ੍ਾਲ(pranītapāla,sukhajītapāla)

29.

ਤੇਜ(tēja)

ਗੁਰੂਤੇਜ,ਕੁਲਤੇਜ(gurūtēja,kulatēja)

30.

ਾਿੰਦਰ(dara)

ਸੁਰਜਿੰ ਦਰ,ਤੇਜਰਵਿੰ ਦਰ(sujidara,tējavidara)

[404]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Issues in Word Segmentation of Handwritten Text in Devanagari Script Rohit Sachdevaa, Dharam Veer Sharmab a

Department of Computer Science, M M Modi College, Patiala, India, Department of Computer Science, Punjabi University, Patiala, India

b

Abstract Word segmentation means dividing the word into sub-parts to extract identifiable units. In handwritten word recognition system, segmentation is a significant pre-processing step. The Indic scripts (such as Bangla, Devanagari, Gujrati, Gurumukhi, etc.) are cursive in nature, making it a challenging task to segment the word. As character set of Devanagari scripts consists of large number of characters further subdivided into consonants, half consonants compound characters, modifiers conjunct consonants and vowels etc. that makes it difficult to segment word Devanagari script. As every person has different method of writing, so it is challenging task to fragment the word into three zones- upper, middle, lower as compare to printed version. The variance in handwritten word is due to: different size of letters, spacing between characters, writing at different angle etc. At times, while writing, the characters tend to be merged, overlapped which adds to complexity of segmentation of these words. This is a general problem, which also occurs very often in other Indic Scripts. Due to the curved nature and inclination to write word at an angle, it is painstaking task to the segment the words of Devanagari script. This paper discusses the issues and problems related to segmentation of handwritten words written in Devanagari Scripts.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: OCR; Handwritten Word Segmentation; Devanagari Script

1. Introduction Text recognition is heavily dependent on appropriate segmentation of text into lines, words and then individual characters or sub-characters as well as feature extraction and classification of the individual characters. Correct segmentation is highly important because incorrect recognition happens if the segmentation has not been done properly. Cursive nature of a script makes the process more assiduous. The structural characteristics of Devanagari character set, curve nature and different ways of writing words makes handwritten word segmentation in Devanagari script onerous job. The core objective of this paper is to identify the complexities involved in the segmentation of hand written words of Devanagari script, to devise suitable solutions for the same in near future. The rest of the paper is organized as follows: section 2 discusses the work done in this area, section 3 describes the properties of Devanagari script, section 4 briefly discusses the method of segmenting hand written words, section 5 deals with the issues faced in segmentation of handwritten word, the papers findings have been summarized in the section 6 and references are given at the end. 2. Previous Work and Related Research Researchers have applied various different approaches for the segmentation. A lot of research has taken place in the field of segmentation of handwritten text in Arabic, Chinese, Japanese and Roman scripts and various methodologies have been postulated by various researchers in handwritten text recognition (Wang et al, 2000; Verma, 2003; Ariki,1995; Plamondon, Srihari, 2000; Chin, Han, 2000; Kim, Bang, 2000).

___________________________ * Corresponding author. Tel.: +91 8146 700347. E-mail address: [email protected].

ISBN: 978-93-82288-63-3

Sachdeva and Sharma/ COMMUNE-2015

During the recent years number of researchers has tried to recognize the off-line handwritten text of Indian scripts. A functional system exists for recognition of handwritten numerals and characters of off-line Bangla script Roy et al, 2005). For Indian postal automation a system has also been developed for unconstrained Bangla handwritten word recognition (Pal et al, 2006). Some research work has been reported for other Indic scripts including Gurmukhi Sharma, (Lehal, 2006), Oriya (Tripathy, Pal, 2006) and Devanagari (Gargl et al, 2010; Ramteke, Rane, 2012; Palakollu et al, 2012). N. K. Garg et al. (Gargl et al, 2010) suggested a technique for line segmentation of handwritten Hindi text. It is mutated form of their earlier proposed method with some presumptions associated with the height of consonant, lower modifiers and maximum height of consonant and skew between the two lines in a text. Their system was depends on detection of head line, base line and contour following method. A. S. Ramteke et al. (Ramteke, Rane, 2012) proposed a method which is implemented in Matlab for segmentation of offline handwritten Devanagari Script. The accuracy of segmentation for this method depends upon the proper writing i.e. proper association of characters with Shirorekha, decent space between words and characters, non-overlapping or characters. The accuracy of segmentation for character, word and numerical achieved by system is 97%, 98% and 100% respectively. Results for broken characters are not so good. S. Palakollu et al. (Palakollu et al, 2012) proposed new technique for segmentation of line and overlapping characters of Handwritten Hindi text. Initially the text is segmented into lines then lines are segmented into words. Later from these words, headline detected and transformed as straight line. Headline and base line is detected by reckoning the average line height and based on it. Before segmentation, skew detection and correction is needed for straightening the image for proper segmentation and feature extractions. N. K. Garg et al. (Gargl et al, 2010) introduced procedure for segmentation of handwritten Hindi text which is based on structural approach. The accuracy of segmentation for proposed procedure gives good results if headline and base line meticulously determined. The accuracy of segmentation for line, word, consonants, ascenders, descenders achieved by procedure is 91.5%, 98.1%, 79.12%, 95.5% and 82.6% respectively. Results for large skewed and touching lines are not great. Munish Kumar et al. (Kumar et al, 2014) illustrated a method for to identify and segment of touching characters in Gurumukhi script by using water reservoir method. By using reservoir base area point, the accuracy for character segmentation achieved by method is 93.51%. Results for broken and overlapped characters are not so virtuous. Preeminent mechanisms for skew detection encompass Correlation Method, Hough Transform, Fourier method, Projection profile, Historate Method, Nearest Neighbor (Postl, 1986; Hashizume et al,1986; Yan, 1993; Baird, 1987; Srihari, Govindaraju, 1989; Hinds et al, 1990; Sharma et al, 2009). Sharma et al. (Sharma, Lehal, 2009) gave the sturdy method, which detect the skew and corrected the isolated words of machine printed Gurumukhi papers. According to authors, if isolated words have straight headline then it is not considered as skewed but when length of headline is less than a threshold value then it is considered as skewed word and need to be corrected. Scripts in which word contain headline used to connect characters, this method is equally adequate for machine printed documents. This method is adequately practiced on Bangla, Devanagari, and Gujarati script words as these have the same structural characteristics as Gurmukhi Script. 3. Characteristics of Devanagari Script Devanagari is part of Brahmi family of scripts of India, which are used in Nepal, Tibet, and south East Asia. Devanagari script has 34 consonants, 12 vowels, 14 modifiers of vowels. It also has compound characters, which are, compose by combining two or more basic characters. As Devanagari script is phonetic and syllabic script so words are written literally as they are speaks. Another typical feature of Devanagari is the existence of a horizontal line on the top of all characters. This line is known as headline or Shirorekha. This line divides the word into two parts. The upper part consists, mostly, of vowels (matras) and modifiers. 3.1

Data Collection:

To analyse the complexities of the handwritten text a form is designed which contains 25 words. Databases constructed by taking data from 200 users it contain a set of 5000 isolated words. These forms are scanned at 300 dpi resolution. In the database, words contained different sizes of characters, touching characters, overlapped characters and slanted characters. In pre-processing step normalization is performed on the data set. Fig. 1 shows some samples of the handwritten words collected from different users.

Fig. 1 Example of handwritten words in Devanagari script

4. Handwritten Word Segmentation in Devanagari Script Handwritten words can be classified into following categories:

[406]

Sachdeva and Sharma/ COMMUNE-2015

a. b. c. d. e.

Words with headline (Fig. 2.a) Words with slanted headline (Fig. 2.b) Words with broken headline (Fig. 2.c) Words without headline (Fig. 2.d) Words with uneven headline (Fig. 2.e)

For segmenting words to individual characters, headline detection is crucial step. In handwritten text, due to skew in the word of writing style of the user, headline may not be a straight line. By using Horizontal Projection Profiles (HPPs) existence or absence of headline can be identified. If the headline is not determined because of its absence then vertical Projection Profiles (VPPs) are constructed to determine gaps for segregation. Figure 3 shows some example of handwritten word’s headline detection.

Fig 3 shows the headline detection using HPPs

If headline is determined, then the whole image is prorated horizontally into upper zone and middle-lower zone. Upper zone part is over the headline and the middle-lower zone under the headline. The determined headline is stripped, to detach the characters in upper and middle-lower zone. This makes gaps in the characters of the word, which are segmented by constructed Vertical Projection Profiles (VPPs). For upper zone characters, deal with the area just over the headline as the bottom of upper zone and creates VPPs. Each gap in the VPPs shows the cut point for segmentation of upper zone characters. Upper zone contains all the vowels modifiers (matras) above the headline. Middle-lower zone is further divided into two sub zones: middle zone and lower zone. Middle zone characters are determined using Vertical Projection Profiles (VPPs), constructed by considering the area just under the headline as the top starting point. Words may contain some broken characters to avert over-segmentation a distance threshold value is used. This step gives us the number of columns constructed in the word, which helps in relating characters in the upper and lower zone with characters in the middle zone. Middle zone characters segmentation is accomplished in three stages. Under first stage the word is segmented, words which are under-segmented are more segmented to the maximum possible limit in the second stage and in the third stage over segmentation is managed resulting from the first stage, which mainly occurs because of broken characters in handwritten text. From the lower part, just under the headline, again constructed Horizontal Projection Profiles and from the bottom move upwards to locate each gap in the HPPs. Existence of any gap in the HPPs shows existence of characters in the lower zone. If no gaps are located to a threshold then either the lower zone characters are overlapped or connected with the middle zone character or not present in lower zone. Figure 4 shows word divided into three zones.

Fig 4 Word divided into three zones

[407]

Sachdeva and Sharma/ COMMUNE-2015

5. Word Segmentation of Handwritten Text in Devanagari Scripts- Issues and Problems 5.1. Problem of Skewed Words: For proper segmentation of a word, the image should be free of any skewness. Some examples of skewed words written in Devanagari script are shown in Figure 5.

Fig 5: Shows skewed words written in Devanagari script

There are several proposed approaches as options for skew angle detection of document images. All the approaches need a rich text area to be present in order to work accurately. Rich text areas carry a well-known characteristic structure, one or more (separate) lines of printed or handwritten words, sharing a common direction. The method proposed in (Hinds et al, 1990) has been used for skew angle detection and correction. Results of applying this algorithm on skewed images are shown in table 1. Table 1: Comparison of word before and after skew correction

Skewed Word

After Skew Correction

5.2. Headline Detection and Removal for segmentation For segmenting words to extract individual characters, headline detection is necessary. Existence or absence of headline can be identified by using Horizontal Projection Profiles (HPPs) as discussed in section 4. This can also be done by locating the rows which have maximum number of black pixels in a word. After the detecting headline we stripped it so as to divide the word vertically into two parts. Table 2: Comparison of word before and after skew correction

Skewed Word Headline detection

After Skew Correction Headline detection

5.3. Overlapping, Connected, Merged or Fused and Broken Characters Segmentation Middle zone contains the overlapped, connected, merged or fused and broken characters. Segmentation of these types of words is a challenging job. Figure 6 shows some examples of these types of words:

(a) Fig 6: (a) Words with overlapping characters

[408]

Sachdeva and Sharma/ COMMUNE-2015

(b)

(c)

(d) Fig 6: (b) Words with connected characters (c) Words with fused or merged characters (d) Words with broken characters

Presence of these types of complexities makes the segmentation task more challenging as shown in Figure 7.

Fig 7.1 Overlapped characters returned two characters as single character (under-segmentation)

Fig 7.2 Connected characters returned two characters as single character (under-segmentation)

Fig 7.3 Fused characters return fused characters as a single character (under-segmentation)

Fig 7.4 Broken Characters returns single character in two characters (over-segmentation)

6. Conclusion If the words are skewed then with the help of skew detection and correction algorithm, the headline detection becomes easy. Then the word is segmented into three zones using Horizontal Projection Profiles (HPPs). Zone wise characters are recognized by using Vertical Projection Profiles (VPPs). Upper zone and lower zone character identification is uncomplicated as compare the characters in middle zone. Middle zone characters segmentation becomes more complex if it contains the overlapped, connected, merged or fused and broken characters words. This eliminates the vertical gaps between the characters and further segmentation is done using cut-classification or recognition based segmentation References Wang, X., Govindaraju, V. and Srihari, S. N., 2000. Holistic Recognition of Handwritten Character Pairs, Pattern Recognition, Vol. 33, pp. 19671973. Verma, B., 2003. A Contour Code Feature Based Segmentation for Handwriting Recognition, in the proceedings of 7th International Conference on Document Analysis and Recognition, pp. 1203 – 1207. Ariki, Y. and Mot, Y., 1995. Segmentation and Recognition of Handwritten Characters using Subspace Method, in the proceedings of 3rd International Conference on Document Analysis and Recognition, Vol. 1, pp. 120-123. Plamondon, R. and Srihari, S. N., 2000. On-Line and Off-line Handwritten Recognition: A Comprehensive Survey, IEEE Transaction on Pattern Analysis and Machine Intelligence, Vol.22, pp.62-84. Chin, Z. and Yan, H., 2000. A Handwritten Character Recognition using Self-organizing Maps and Fuzzy Rules”, Pattern Recognition, Vol.22, pp. 923-937.

[409]

Sachdeva and Sharma/ COMMUNE-2015 Kim, K. and Bang, S.Y., 2000. A Handwritten Character Classification using Tolerant Rough Set, IEEE Transaction on Pattern Analysis and Machine Intelligence, Vol.22, pp.923-937. Roy, K., Pal, U. and Kimura, F., 2005. Recognition of Handwritten Bangla Characters, in the proceedings of 2nd International Conference on Machine Intelligence, pp.480-485. Pal, U., Roy, K. and Kimura, F., 2006. A Lexicon Driven Method for Unconstrained Bangla Handwritten Word Recognition, in the proceedings of 10th International Workshop on Frontiers in Handwriting Recognition, pp.601-606. Sharma, D. V. and Lehal, G. S. 2006. An Iterative Algorithm for Segmentation of Isolated Handwritten Words in Gurmukhi Script, in the proceedings of 18th International Conference on Pattern Recognition, Vol. 2, 1022-1025. Tripathy, N. and Pal, U. 2006. Handwriting Segmentation of Unconstrained Oriya Text, Sadhan, Vol. 31, pp. 755–769. Garg, N. K., Kaur, L. and Jindal, M. K., 2010. A New Method for Line Segmentation of Handwritten Hindi Text, in the proceedings of 7th International IEEE Conference on Human Technology: New Generations (ITNG), pp. 392- 397. Ramteke, A. S. and Rane, M. E., 2012. Offline Handwritten Devanagari Script Segmentation, International Journal of Scientific and Technology Research, Vol. 1, Issue 4, pp. 142-145. Palakollu, S., Dhir R. and Rani, R., 2012. Handwritten Hindi Text Segmentation Techniques for Lines and Characters, in the proceedings of the World Congress on Engineering and Computer Science, Vol. 1, pp. 640-644. Garg, N.K., Kaur, L., and Jindal, M. K., 2010. Segmentation of Handwritten Hindi Text, International Journal of Computer Applications (0975 – 8887), Volume 1 – No. 4 pp. 19-23. Kumar, M., Jindal, M. K. and Sharma, R. K., 2014. Segmentation of Isolated and Touching Characters in Offline Handwritten Gurmukhi Script Recognition, International Journal of Information Technology and Computer Science, Vol. 02, pp 58-63. Postl, W. 1986. Detection of Linear Oblique Structures and Skew Scan in Digitized Documents, in the proceedings of International Conference on Pattern Recognition, pp. 687-689. Hashizume, A., Yeh, P. S. and Rosenfeld, A., 1986. A Method of Detecting the Orientation of Aligned Components, Pattern Recognition Letters, Vol. 4, pp. 125-132. Yan, H., 1993. Skew Correction of Document Images Using Interline Cross-correlation, Computer Vision Graphics and Image Processing: Graphical Models and Image Processing, Vol. 55, Issue 6, pp.538-543. Baird, H. S., 1987. The Skew Angle of Printed Documents, in the proceedings of Conference on Photographic Scientists and Engineers, pp. 14-21. Srihari, S. N. and Govindaraju, V., 1989. Analysis of Textual Images using the Hough Transform, Machine Vision and Applications, Vol. 2, pp. 141153. Hinds, S. C., Fisher, J. L. and Amato, D. P. 1990. A Document Skew Detection Method using Run-length Encoding and the Hough Transformh transform, in the proceedings of International Conference on Pattern Recognition, pp. 464-468. Sharma, D. V., Gupta, S. and Beri, P., 2005. Skew Angle Detection and Correction of Hand Written Gurmukhi Words using Historate Method, in the proceedings of the International Conference on Cognition and Recognition, pp. 22-24. Sharma, D. V. and Lehal, G. S., 2009. A Fast Skew Detection and Correction Algorithm for Machine Printed Words in Gurmukhi Script, presented in an International Workshop on Multilingual OCR, published by ACM, article no 15.

[410]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Remote Monitoring of Water Pollution in Dal Lake using Wireless Sensor Networks in Realtime Sofi Shabir, Roohie Naaz National Institute of Technology, Srinagar, 190006

Abstract Pollution monitoring in water is an important part of environment monitoring. The population boom, high urbanization levels and the birth of new industries are producing adverse effects on water bodies and leading to pollution of water and on environment in general. Due to the constraints of the natural conditions and time and temporal factors, the traditional monitoring methods have some limitations. In recent years remote sensing technologies were widely applied in the investigation and monitoring of various pollutions in aquatic environments. Field investigation is reliable but too expensive and time consuming. Satellite remote sensing have the advantages of region scale coverage and moderate timeliness but restricted by the factors of heavy mist, low image resolution and prohibitive cost. For better accuracy and spatial coverage Wireless Sensor Networks can be promising approach with real time monitoring of water pollution and aquatic life. In this paper we analyse the different techniques used in water pollution monitoring and propose a cost effective new technique for the monitoring of water bodies based on Wireless Sensor Networks with transferring of the data collected in real time and viewing of the same via a network. The GSM model of the same has been tested in the Lab and is working satisfactorily.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Wireless Sensor Network ; Dissolved Oxygen ; Wi-Max Technology;Gateway; GSM;Pollution, Environment

1. Introduction Water is the essential resources for all the living organics of the earth to survive. The quality of life on earth is linked inextricably to the overall quality of the environment. Nevertheless, water is suffering from various pollutions due to thousands of reasons. The major reasons are population boom; high urbanization levels and the birth of new industries producing adverse effects on water bodies and leading to pollution of water and on environment in general. Natural phenomenon such as volcanoes, algal blooms, storms and earthquakes also cause major change in water quality and ecological status of the water. Since water environment monitoring is an important part of environment monitoring, therefore it is of great importance to develop an effective water pollution monitoring system. Due to the constraints of the natural conditions and time and temporal factors, the traditional monitoring methods have some limitations. Field investigation is reliable but too expensive and time consuming. In recent years remote sensing technologies were widely applied in the investigation and monitoring of various pollutions in aquatic environments. But they are also not accurate and limited by many factors especially they fail to provide the realtime monitoring. A new approach is the need of the hour and one of the promising technologies is the wireless sensor networks. 2. Traditional approaches The traditional water inspection method is called offline inspection approach (Qingbo and Jianzhong, 2006). The inspectors collect sample water from the monitored area and bring back to the laboratory for analysis. With this method, 

Corresponding author Tel.: +91 9419 009971. E-mail address: [email protected]

ISBN: 978-93-82288-63-3

Sofi and Naaz/COMMUNE – 2015

the inspection circle is long sometimes it takes a few days to get the result, moreover the analysis result is limited to sampling area and it is unable to carry out realtime monitoring in a big range (Jung and Lee, 2008). The different chemical parameters which are essential indicators of pollution level of the water body are given in table below. Each parameter is having its threshold value in water which should not be allowed beyond a certain limit. For this purpose samples are taken to lab and the analysis is done. Chemical Parameters

Reason for the Analysis

Temperature

Temperature can exert great control over aquatic communities. If the overall water body temperature of a system is altered, an aquatic community shift can be expected.

pH value

pH is an indicator of the existence of biological life as most of them thrive in a quite narrow and critical pH range. DO is essential for aquatic life. A low DO (less than 2mg/l) would indicate poor water quality and thus would have difficulty in sustaining many sensitive aquatic life. Conductivity indicates the presence of ions within the water, usually due to in majority, saline water and in part, leaching. It can also indicate industrial discharges.

Dissolved Oxygen (DO)

Conductivity

3. Remote Sensing The development and maturation of remote sensing technology has made a great breakthrough on testing for the environmental pollutants and has solved dozens of practical problems. The remote sensing technology has wide monitoring range; high speed, low cost and being convenient to make long term dynamic monitoring as compared to traditional approach (Sun et al, 2010). 4. Polarized Remote Sensing Using multi-band, multi-date and hyper-spectral remote sensing data, the ability to identify surface features increases. It is found that the angle information has made great influence and contribution on the distinguish and classifying of remote images (Fenghui et al, 2009) which is called three dimensional spectral characteristics of surface features in 2π space. The polarized remote sensing information can show richer content in expressing dark objects, it provides a new effectual measure for water remote sensing (Yu et al, 2010). Polarization makes great sense of expressing the nature of water and the nature of material in water. Thus polarized remote sensing technology can be used as a kind of supplementary means so that it could play its role in water quality monitoring. Satellite remote sensing have the advantages of region scale coverage and moderate timeliness but restricted by the factors of heavy mist, low image resolution and prohibitive cost and accuracy. For better accuracy and spatial coverage Wireless Sensor Networks can be promising approach with realtime monitoring of water pollution and aquatic life. 5. Wireless Sensor Networks (WSN) Advances in electronics and wireless communications have enabled a new evolution in wireless sensor networks. Over the last 5 years WSN have attracted a great deal of interest due to their cost-effectiveness, ability to perform multiple functions simultaneously and make decision based on information gathered from various sensing elements placed at different locations. A wireless sensor network consists of a number of nodes which are low costs, with wireless communication, sensing, and data processing capabilities which can be distributed across a geographical area. A sensor node consists of information collection module, responsible for monitoring the area of information collection and data conversion; information processing module, responsible for controlling the operation of the sensor nodes, storage and processing their own data collection and data sent by the other nodes; information exchange module, with other sensor nodes for wireless communication, exchange control messages and send and receive data collection; system power module for the sensor nodes to provide the energy required to run, usually in a miniature battery. The sensor part of the node is responsible for the physical sensing of the parameter such as temperature, ph value, dissolved Oxygen etc. The parameter list as given in the table above can be more to depending upon the availability of the sensors. Only the sensor part is changed for the WSN node rest of the components remains same as shown in fig 1. remains same.

[412]

Sofi and Naaz/COMMUNE – 2015

Fig 1: Sensor Node Architecture

A wireless sensor network is kind of ad hoc network which is easily configured, without infrastructure, such as cables or structures. Since a sensor node is low priced and can be placed in a large area of water. The cost of deployment is less. Also many sensor nodes could monitor the same area at the same time through compressed configuration so that the inspection result will be more accurate through redundancy data analysis, the inspection period is short and the inspection is in realtime. Proposed Wireless Sensor Network for water monitoring Architecture:

Fig 2: Proposed Architecture

The architecture of the proposed water monitoring system is shown in fig. 2. Wireless sensor nodes are deployed at various locations of the water body depending on the need and parameters to be monitored. The nodes can have ability of monitoring one or many parameters at a time. The topology for the network is cluster based. Each group of the nodes [413]

Sofi and Naaz/COMMUNE – 2015

is having a cluster head. All the nodes in the cluster are communicating with the cluster head for onward transmission to the gateway. The cluster head can be static or dynamic. In static case the cluster head is predefined and it knows the complete topology of the network beforehand. In case of the dynamic cluster head each node may get a chance to become cluster head depending upon the energy level of each node. The node with higher energy level will become the cluster head. In the later case the network life will be improved (shabir and Naaz, 2010). In our case we are using a topology based on static cluster head. The cluster head will be without the sensing part and having a higher energy level than the individual nodes [dummy node] and having higher transmitting range with the capability for the communicating with the nodes using Zigbee technology and with the gateway using Wimax technology (Silva et al). The gateway in turn can communicate with the nearest base station and the base station in turn is communicating with WSN Data Server. The server is having the necessary software to analyse the data received and then the processed data is stored in a database. The database is then connected to the internet and the users can access the data in realtime irrespective of the place. Since the monitoring is not only required at one place of the water body but we need to establish many such clusters with a separate gateway for each cluster. The base station can be one or many depending on the area of the water body and the range of the base station. The nodes may directly communicate with the base station or the data may be routed through the other nearby nodes in case the sending node is out of range. Again the gateway may communicate the data to be sent to the base station via another nearby gateway in case it is out of range. 6. Conclusion Due to the constraints of the natural conditions and time and temporal factors, the traditional monitoring methods have some limitations. Satellite remote sensing have the advantages of region scale coverage and moderate timeliness but restricted by the factors of heavy mist, low image resolution. For better accuracy and spatial coverage, Wireless Sensor Networks can be promising approach with realtime monitoring of water pollution and aquatic life. The proposed architecture can use the available technologies of communication with the power of physical monitoring of the environment around us by the tiny WS nodes. The architecture has the limitation that it can be useful to medium sized water bodies like lakes etc. only. There are many challenges which include: deployment, topology control, battery recharge, device design etc. All these can be taken for future work. References Jiang Peng, Huang Qingbo and Wang Jianzhong Research on Wireless Sensor Networks Routing Protocol for Water Environment Monitoring 07695-2616-0/06 2006 IEEE. JY Jung and JW Lee, “ZigBee Device Access Control and Reliable Data Transmission in ZigBee Based Health Monitoring System,” ICACT 2008. 10th International Conference on Advanced Communication Technology, vol. 1, pp. 795-797, February 2008. Sofi shabir and Roohie Naaz “ Dummy Node based Design of Energy Efficient Wireless Sensor Network” IET 2010 Steven Silva , Hoang N ghia Nguyen , Valentina Tiporlini and Kamal Alameh Electron Science Research Institute Web Based Water Quality Monitoring with Sensor Network: Employing ZigBee and WiMax Technologies Yang Yu, Xinhua Li and Liang Chen “The Application of Wireless Sensor Networking in Environmental Monitoring Based on LEACH Protocol”IEEE 2010 Zhang Fenghui, Zhou Huiling and Zhou Xiaoguang, “A Routing Algorithm for ZigBee Network Based on Dynamic Energy Consumption Decisive Path,” CINC’09. International Conference on Computational Intelligence and Natural Computing, vol. 1, pp. 429-432, June 2009. Zhongqui Sun, Yusheng Zhao and Shaoping Li “Research on polarized remote sensing of monitoring of water pollution “ IEEE 2010

[414]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Machine to Machine (M2M) Control & Communication for Internet of Things (IoT) using DTMF G. Mohiuddin Bhat, Rouf-ul-Alam Bhat, Naazira Badar, Malik Rabaie Mushtaq, Afzan Hussain Hamdani Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar, India

Abstract Dual Tone Multiple Frequency (DTMF) is a low cost, global wireless control technology used in diverse application areas. Although a lot of work has been done on the applications based on DTMF, major portion of the research and development has been carried out to develop systems that involve a manual data entry at the transmitter side and an automatic response at the receiver side e.g. the Interactive Voice Response (IVR) System that creates a lot of scope for human to machine (H2M) interaction. With the advent of Cloud centric Internet of Things (IoT) technology, there is a need to design and develop a low cost, globally accessible machine to machine (M2M) interaction technology. In this paper, a system is proposed and implemented that makes the DTMF technology two way interactive and autonomous so that it can be used for M2M IoT market. Although a class of data sources can be used, the experimentation has been performed through a test bed where an image processing application is automatically generating control signals to be relayed over Global System for Mobile Communications (GSM) to a remote system thus emulating an M2M system

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: IOT; M2M; GSM; DTMF; H2M

1. Introduction Kevin Ashton was the first person to coin the term Internet of Things in 1999. He introduced this term in reference to the concept of supply chain (K. Ashton, 2009) management. However, this definition has been more comprehensive in the past decade covering broader range of applications such as health care, utilities and transport (H. Sundmaeker et al., 2010). Although with the evolution of technology, the definition of ‘Things’ has changed, the goal of prime importance still remains to make a computer sense information without the aid of human intervention. Internet of Things (IoT) is expected to be a huge leap radically evolving the current Internet into a network of interconnected objects that not only takes information from the environment via the sensing module and interacts with the physical world via the actuation, command and control modules, but also uses the status quo Internet standards to provide various services like information transfer, communications, analytics and applications. Strongly propelled by the prevalent open wireless technologies like Bluetooth, radio frequency identification (RFID) and Wi-Fi as well as embedded sensor and actuator nodes, IoT has stepped out of its initial stages and is on the verge of transforming the current static Internet into a fully integrated Future Internet (J.Buckley, 2006).The propagation of mobile Internet has resulted in ubiquitous mobility and nationwide coverage. The cost of broadband data service provided by today's advanced wireless networks is significantly less than in the past due to extensive standardization (Surobhi et al., 2014). Machine to machine (M2M) broadly refers to any technology that helps networked devices in exchanging information and makes these capable of performing tasks without the manual human assistance. M2M is an indispensable part of the IoT and is very beneficial for the industry and business in general.

_________________________________ * Corresponding author. Tel.: +91 9596 088846 E-mail address: [email protected].

ISBN: 978-93-82288-63-3

Bhat et al/ COMMUNE-2015

The applications of M2M spread over a large extent in areas like industrial automation, logistics, Smart Grid, Smart Cities, health and defence mostly for monitoring but also for control purpose. Technology is one of the major driving forces of M2M.The growing semiconductor industry provides better miniaturization along with improved yield, thereby continuing to reduce cost and power consumption per chip. The coverage of wireless networks can be further extended employing the use of peer-to-peer communication, relays and small cells while drastically reducing the cost per transmitted bit. The economic reasons are also fervent motivations for the wireless industry to adopt M2M. The deteriorating voice revenue has put operators are under massive pressure to introduce new services to fill this revenue gap. Cloud computing, M2M and application stores lead the list of potential revenue-generating services (Geng Wu et al.,2011).Based on the definition given by the European Telecommunications Standards Institute (ETSI), M2M architecture can technically refer to a variable number of communicating machines. However, it is generally accepted that M2M principles are valid particularly well for networks where a large number of machines is used, even up to the estimated 1.5 billion wireless devices of the future. This means that when an M2M application is discussed, it is presented on a global scale with a profusion of centrally coordinated sensors (M.J. Booysen et al., 2009). 2. Background DTMF is a very reliable means of signaling and is being used largely for global communication and control. Although a number of protocols have been developed for wireless communication, these are limited by the short range (Nasim uz zaman Chowdhury and Md. Khaled, 2013). Employing the use of DTMF, this limitation can be overcome. DTMF can be transmitted over telephone lines as well as over the Internet. Whenever a key is pressed on the DTMF encoder, a DTMF tone is generated and transmitted. This DTMF tone is decoded at the receiving end and used for practical applications related to communication and control. In case of mobile phones, DTMF tones can be generated only after the connection is established. Most of the applications employing DTMF for control use a GSM phone connected to a cellular network for transmitting signals. The controls are provided by manual pressing of the keypad by a human. These DTMF-based systems have the following drawbacks: ● The speed of control is limited by the rate at which a human can press the buttons of the DTMF keypad. ● These designs are costly. ● These designs require more power. In this research paper, we have focused on the proposed system to make both the encoder and decoder modules of our DTMF - controlled robot fully automated and low cost employing the M2M technology unlike the H2M based control architectures of other robots. The devices that constitute the Internet of Things (IoT) have a variety of low power microcontrollers, sensor networks and communication protocols at the core. These underlying technologies operate in a fast and dynamic environment providing services like tele-monitoring, M2M communications and motion control. Microcontrollers are monitoring and controlling power close to the load point, providing better device management. When low power design architectures combine with manufacturing designed to be power sensitive, the total power budget and costs are dramatically reduced. Moreover, with the use of a GSM based technology, we have achieved a global control. 3. Problem Definition A lot of work has been done on DTMF technology and its efficient implementation, but most of the advancement in the DTMF has been done at the decoder end. Human intervention is still required at the encoder end which restrains the system from being a fully automatic M2M system. Moreover, the conventional systems are costly and have large power requirements. 4. Proposed Solution – Design And Implementation In this paper, a system is proposed and implemented that makes the DTMF technology two way interactive and autonomous so that it can be used for M2M IoT market. The block diagram in Fig. 1 describes the overall functionality of our proposed design. To emulate an M2M device, an image processing application has been designed that senses color information using the image acquisition device and converts it into a set of codes. The acquired data is fed to a computing device with MATLAB installed on it. MATLAB has been programed to generate serial data corresponding to the input. The serial data are interfaced with the processing module consisting of an Atmega328 microcontroller based Arduino UNO board using an RS-232 cable. The Atmega328 has been programed to do serial to parallel conversion the parallel data act as control signals for the DTMF encoder module corresponding to which different DTMF tones are generated which are fed to the communication network (GSM) through a communication module. The communication module acts as an interface between the encoder module and a GSM mobile handset that has been used as a relaying element such that these tones reach the receiving station over the GSM link. The backbone of encoder module is the DTMF encoder IC-UM91214B and a CMOS analog switching ICCD4066BE. Two CD4066BE ICs have been used for producing a total number of eight DTMF tones. The parallel data [416]

Bhat et al/ COMMUNE-2015

generated by the processing module are the controlling inputs for the switches of these ICs. The outputs of these switches are given to DTMF encoder IC which generates a particular tone for a particular input combination of low and high group frequency which is the basis of DTMF operation. The frequency combination table is shown in Table 1. Table 1. Frequency combination table High group frequencies

1209

1336

1477

1633

697

1

2

3

A

770

4

5

6

B

852

7

8

9

C

941

*

0

#

D

Low group frequencies

Fig. 1. Block diagram of the proposed architecture

The communication module consists of a microphone jack that acts as a tone-handset interface and has been primarily designed for taking the input from condenser or capacitive microphones. The following two design approaches have been considered for the same: ● Use of acoustic coupling: In this approach, the output from the encoder is directly fed to a speaker which is acoustically coupled with a microphone. This approach resulted in an advantage of providing a better isolation but had lesser speed. The maximum number of tones that could be transmitted per second was two.

Fig. 2. Acoustic Coupling Device

● Use of pi attenuator: In this approach, the output of encoder is fed to microphone input through an attenuator. The attenuator is needed because the incoming signal from encoder is of the order of hundreds of millivolts but the maximum input that can be fed to the microphone jack is of the order of few hundred microvolts.This approach although lacked in proper isolation but resulted in better speed and performance than acoustic coupling. Hence a pi type attenuator has been used in the prototype with a low pass filter for filtering out the high frequency noise. The prototype M2M transmitter for our proposed design is shown in Fig. 4. While as The prototype M2M receiver for our proposed design is shown in Fig. 5.

[417]

Bhat et al/ COMMUNE-2015

5. Results Out of all the possible frequency combinations shown in Table 1, only eight have been used for our prototype. For each of the eight combinations, a DTMF tone was generated. The resultant tones were viewed on the Digital Storage Oscilloscope as shown in Fig. 6.

Fig. 3. Attenuator

Fig. 4. Prototype M2M transmitter

Fig. 5. Prototype M2M receiver

(a)

(b)

[418]

Bhat et al/ COMMUNE-2015

(c)

(d)

(e)

(f)

(g)

(h)

Fig. 6. (a) tone produced corresponding to Frequency Combination 1209 Hz with 697 Hz, (b 1336 Hz with 697 Hz, (c) 1477 Hz with 697 Hz, (d) 1209 Hz with 770 Hz, (e) 1336 Hz with 770 Hz, (f) 1477 Hz with 770 Hz, (g) 1209 Hz with 852 Hz

For each of the tone shown in Fig. 6, the resultant tone amplitude is 500-800mV. 6. Conclusion The paper presents the proposed architecture for the M2M based DTMF encoder design. The results clearly reveal that the DTMF tones can be faithfully generated without manually pressing the keypad of mobile phone eliminating the human intervention. Two design approaches have been considered for interfacing the tone with the microphone. The piattenuator approach resulted in better performance and has been implemented in the final design. The proposed architecture should help in shifting the DTMF encoder design from H2M technology to M2M technology for the IoT. References K. Ashton, 2009. That ‘‘Internet of Things’’ thing, RFiD Journal. H. Sundmaeker, P. Guillemin, P. Friess, S. Woelfflé,March 2010. Vision and challenges for realising the Internet of Things, Cluster of European Research Projects on the Internet of Things—CERP IoT. J. Buckley, 2006. The Internet of Things: From RFID to the Next-Generation Pervasive Networked Systems, Auerbach Publications, New York. N.A. Surobhi, A. Jamalipour, 2014.A Context-Aware M2M-Based Middleware for Service Selection in Mobile Ad-Hoc Networks, IEEE Transactions on Parallel and Distributed Systems, p. 3056-3065. Geng Wu,S. Talwar, K.Johnsson,N. Himayat , K.D. Johnson,April 2011. M2M: From mobile to embedded internet, Communications Magazine, IEEE, p.36-43. M.J. Booysen, J.S. Gilmore1, S. Zeadally, G.J.van Rooyen,Feb. 2012. Machine-to-Machine (M2M) Communications in Vehicular Networks/ KSII Transactions on Internet and Information Systems, p.529-546. Nasimuzzaman Chowdhury, Md. Khaled Hussain, 2013. M2M: GSM Network for Robots using DTMF, The Global Journal of Researches in Engineering, p.23-29.

[419]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Framework for Web Security Using Multimedia Password Authentication Manzoor Ahmad Chachoo, Farah Fayaz Quraishi, Summera Ashraf * Department of Computer Sciences , University of Kashmir , Srinagar , Kashmir, India

Abstract Various studies over a long period give sufficient proof that humans are better at remembering images compared to text. By taking into the account the above fact , the team members strived to bring such a system (Multimedia Password Authentication) that is more secure, supports memorability and has a user friendly interface and uses the memory trigger connected to previously seen image. For enhanced security and efficiency, various sensitivity levels viz. Easy, Medium, Hard were introduced. The relation between security and usability is an anomaly that is not grasped in the current systems(mostly alphanumeric).The research is focused on whether the multimedia passwords(images along with audio track and video clips) deal with important issues like memorability and the ever important security factor at the same time while being practical in usage. To aid in remembrance of passwords (if forgot), images are used in place of traditional security questions for enhanced security. In addition, user is allowed to provide for Password Hint, in which text is converted to speech. The converted text which can be referred to as the sound at this point of time provides the user with the convenience of recovering forgotten passwords.Upon usage of the multimedia password authentication system, a significant increase was seen in the performance of the system regarding important benchmarks such as accuracy and speed. This significant increase in various important aspects of the system was seen independent of the time consumption of the user, which usually took longer than alphanumeric ones.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Information Security, Multimedia passwords, Image Click points, Audio Tracks, Video Clips, Sensitivity Levels

1. Introduction Patrick and Dourish (2003-04 respectively)stated that security is fundamentally a human-computer interaction(HCI)problem. The human-computer interaction is extremely important in two ways 1) Usability of the security mechanisms. 2) Interaction of security mechanisms with the user. The usability or efficiency of any given password depends on the authentication system software and how the information regarding that particular password is stored. Alphanumerical passwords are used to meet two quite opposite aspirations of the user. First of all, they must be easily recalled by a user while at the same time they have to be difficult for a hacker to crack. The general perception regarding the alphanumerical passwords is the usage of words and numbers that can be recalled by the user quite effortlessly (S. Chiasson et al.2007, 2008, 2009, E. Stobert et al, 2010 G.Agarwal, 2010) The dichotomy of using a particularly hard to remember password is that user has to write them down somewhere making them susceptible to hackers rather easily (Adams & Sasse, 1999). (G.Agarwal et al, 2010, Pinks, B. and T. Sander. 200,2). Biometric systems which are supposed to relieve the user of the above dilemma have their own disadvantages. (L. Jones et al , 2007, L. O‟Gorman, 2003, A. Jain, et al, , 2006). As compared to the biometric systems, graphical passwords offer or present a more practical alternative and are the focus of this paper. Blonder defined the graphical passwords in 1996.The Image Based Authentication rests on cognitive ability of association based memorization of humans compared with traditional textual passwords(Zhi Li et al, 2005). In addition to images, audio tracks and video clips have been used in the project to increase the remembrance of passwords. *

Corresponding author. Tel.: +91 9797 847507. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Chachoo et al/COMMUNE – 2015

Multimedia Password Authentication System (MPAS) is an alternative to solve the hardships that come to the fore regarding text based or graphical password system. Multimedia passwords are comparatively tougher to crack either by plain guessing or by using search of any kind. The potential click points in the images, audio clips and videos is rather large, therefore the presumable password area of MPAS scheme tend to overtake text based schemes and hence offer a more durable resistance to attacks. This paper tries to show a fresh, more adaptable and a relatively higher secure password system that we have ventured to design and consequently implement. In addition, we compare the traditional password systems to the multimedia authenticated system. The principal research questions that lead to the development of MPAS are stated below: RQ1: What kind of passwords is required by the user for multiple accounts RQ2: How does a user remember passwords for either a single account or multiple accounts? RQ3: What kind of limitations a user faces during password recovery process? 2. Current Authentication Methods The general classification of the current authentication methods can be expressed under following three headings: a) Token- based Authentication(e.g., Credit cards ) b) Biometric based Authentication(e.g., Iris scan, Facial recognition) The most disappointing aspects of the above two authentication methods is that they are quite expensive and demand special instruments. c) Knowledge based Authentication.(Two types—Text based and Picture based) Text Based: Due to their ease of use textual or text passwords are the go-to choice for humans. As the human capacity to remember is rather limited, the passwords chosen are rather easy to remember. The conventional passwords patterns and schemes are susceptible to various attacks. Keeping this fact in mind it is highly essential that tougher passwords are used for important accounts and the habit of changing the passwords regularly should be encouraged. Picture Based:The inherent problem of using an image as a password is that the image can be easily predicted and hence susceptible to be used by unwanted sources. Assumption is that the location of the user’s three gestures is all chosen independently. In practice, this is not realistic; there will often be some pattern. For instance, in the example the user chose to tap on three windows in sequence .(Fig 1)

Fig. 1 If you’ve guessed that the first tap location is over a window, then it would be natural to guess that maybe the next two are on some other windows, too.

2. Smudge Attacks Past researches have shown if someone gets hold of your laptop/phone, they could guess your picture password by looking at the pattern of smudge marks on the touchpad left by your finger oils (Adam J et al, 2010) (Amazingly, they found that the smudge marks remained clearly visible even if the user put the phone in their pocket -you might expect this would wipe the fingerprints off, but nope, they remained visible! An attacker can guess the picture password and significantly reduce the entropy in the password. For instance,

[421]

Chachoo et al / COMMUNE-2015

suppose the attacker gets lucky and there are only 3 smudge marks on the screen. Then there are 3! = 3*2*1 = 6 possible re-orderings of these. The attacker gets 5 tries to guess the picture password. So, in this scenario, the attacker has a 5/6 chance of guessing the picture password correctly before being locked out. That said, this is almost the best possible case for the attacker. 3. Design of Authentication System based on Multimedia Passwords (MPAS) The main objective behind the development of an authentication system based on Multimedia passwords was to providing a security method for any software application offline or online as an alternative to traditional text passwords, whereby a user must remember an image (or parts of an image) in place of a word. They are motivated in part by the well-known fact that people are better at remembering images than words. In MPAS, a supportive Audio and Video signature is provided to increase in remembrance of passwords, thus enhancing efficiency.(Fig 2) Further the security of this system is increased by providing the user with number of levels for password sensibility which are: a)

Level EASY: In this level, users are allowed to keep click points on Image or combination of click points on same or multiple Images, as their password. b) Level MEDIUM: In this level, users in addition to image click points, are allowed to keep Audio too , as their passwords i.e., users can keep combination of Images and Audio as their passwords. This ensures better security as well as better memorability. c) Level HARD: In this level, users in addition to image click points and audio, are allowed to keep click points on video, as their passwords. A combination of Images, Audio and Video ensures a Highly Secure authentication system.

Fig 2: Levels in MPAS

User choice is influenced by persuasion, encouraging users to select more random, and thus more difficult to guess click-points. MPAS is Highly Reliable and Highly Flexible. It provides lot of benefits: a) Authentication, Authorization and Access Control. b) The system is provided with easy-to-use GUI(graphical user interface) that guides the user in a simple manner to follow the necessary steps without much hindrance. c) Also, To aid in remembrance of passwords(if the password is forgotten), images are used in place of traditional security questions for enhanced security. d) In addition user is allowed to provide for Password Hint, in which text is converted to speech. This sound signature will be used to help the user to recover password.(Fig. 3)

[422]

Chachoo et al/COMMUNE – 2015

Fig .3: Password Recovery 4. Empirical Results and Analysis In the learning Phase, the participants entered the password repeatedly until they achieved correct input submission. The users were allowed to use this system for 10 days, after which they were asked to fill a questionnaire based on the memorability, usability and security of the system. The participants gave the score out of 5 to each of the available password systems.

Fig. 4: 2-D Line Chart depicting the feedback of participants

3. Security An authentication system should give required safety or security for its potential environment otherwise it cannot meet its prime objective. Password brute forcing attack: A system will be susceptible to a brute forcing attack if it does not possess a proper protocol to make sure that passwords are strong and follow the adequate password policy. a)

Dictionary based password attack: As the name suggests, this attack represents the usage of every word in a dictionary as password by the hacker to enter the system through some other user’s account. If the user had used a word present in the dictionary then the attacker is bound to be successful. (Van Oorschot, P.C., S. Stubblebine, 2006) b) Guessing attack :This is a hit or miss type of attack with least success rate among various types of hacks or attacks. In this particular attack the hacker tries to enter the system by presuming that the user password must be based on the personal information of the user e.g. nicknames, etc.

[423]

Chachoo et al / COMMUNE-2015

c)

Spyware attacks: Spyware is a malware that is covertly installed in a system with the intention of scanning information about users by employing key logger or even a key listener. Since the data collected through this procedure is reported to an outside source that needs information about a particular user. While graphical password authentication is going on the hacker tries to obtain the confidential information like a particular username or even a particular image. d) Social engineering attack: This type of attack involves one or more than one interaction of the attacker to obtain sensitive and highly confidential information to infiltrate an organization or computer systems. The hacker asks many relevant questions to achieve his/her goal 4. Conclusion The model proposed above named as MPAS promises to deliver as an easier to use and a memorable authentication mechanism. The MPAS model takes advantage of user’s ability to recognize images and hence has a distinct advantage in terms of usability. MPAS has an edge over other mechanisms as it is definitely more secure. MPAS multiplies the essential workload for hackers by compelling them to first get hold of a particular image selected by the user. This step represents an EASY LEVEL of intrusion for the attackers. Also MPAS makes the attackers acquire image and relative audio tracks chosen by the user(MEDIUM LEVEL).The most HARD LEVEL for an attacker to reach or cross is the one where he/she has to obtain image, audio tracks and video clips and subsequently conduct hotspot analysis (Thorpe, J et al, 2007)on each of the image. This approach has been proven to be very effective at lowering the formation of hotspots, consequently avoid shoulder surfing and at the same time provide the ever needed high security success without compromising usability.

References S. Chiasson, R. Biddle, and P. van Oorschot, 2007 A Second Look at the Usability of Click-Based Graphical Passwords,”Proc. ACM Symp. Usable Privacyand Security (SOUPS). S. Chiasson, A. Forget, R. Biddle, and P. van Oorschot, 2008.Influencing Users towards Better Passwords: Persuasive Cued Click- Points,”Proc. British HCI Group Ann. Conf. People and Computers: Culture, Creativity, Interaction, S. Chiasson, A. Forget, E. Stobert, P. van Oorschot, and R. Bddle, 2009., Multiple Password Interference in Text and Click-Based Graphical Passwords,Proc.ACM Conf. Computer and Comm. Security CCS). E. Stobert, A. Forget, S. Chiasson, P. van Oorschot, and R.Biddle, 2010, Exploring Usability Effects of Increasing Security in Click-Based Graphical Passwords,Proc. Ann. Computer Security Applications Conf. (ACSAC). G.Agarwal,S.Singh and R.S.Shukla, ., 2010, Security Analysis of Graphical Passwords over the Alphanumeric Passwords. Pinks, B. and T. Sander. 200,2, Securing Passwords Against Dictionary Attacks. ACM, CCS, L. Jones, A. Anton, and J. Earp, 2007, Towards Understanding User Perceptions of Authentication Technologies, Proc. ACM Workshop Privacy in Electronic Soc. L. O‟Gorman, 2003, Comparing Passwords, Tokens, and Biometrics for User Authentication, Proc. IEEE, vol. 91, no. 12, pp. 2019-2020, Dec.. A. Jain, A. Ross, and S. Pankanti, 2006, Biometrics: A Tool for Information Security,” IEEE Trans. Information Forensics and Security (TIFS), vol. 1, no. 2, pp.125-143. Zhi Li, QibinSun ,YongLian, and D.D.Giusto,2005, An Association Based Graphical Password Design Resistant to Shoulder Surfing Attack, IEEE International Conference on Multimedia and Expo(ICME) 0-7803-9331-7 ,pp245-248 Adam J. Aviv, Katherine Gibson, Evan Mossop, Matt Blaze, and Jonathan M. Smith, 2010, Smudge Attacks on Smartphone Touchscreen- WOOT. Van Oorschot, P.C., S. Stubblebine, 2006, On Countering Online Dictionary Attacks with Login Histories and Humans-in-the-Loop.ACM Trans. Information and System Security 9(3), 235-258, Thorpe, J. and P.C. van Oorschot, 2007. Human-Seeded Attacks and Exploiting Hots-Spots in Graphical Passwords.16th USENIX Security Symposium, R. N. Shepard, 1967, Recognition memory for words, sentences, and pictures, Journal of Verbal Learning and Verbal Behavior, vol. 6, pp. 156163, Blonder, G.E. 1996. Graphical Passwords.United States Patent 5559961

[424]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Computational Approaches for Emotion Detection and Classification in Textual Data Abid Hussain Wania*, Rana Hashmyb a

Department of Computer Science, South Campus, University of Kashmir, Srinagar, India b Department of Computer Science, University of Kashmir, Srinagar, India

Abstract Men are born with a “human heart” and express their feelings, behavior, experience, physiology, conceptualization, and cognitions through emotions. Internet provides a global platform for people to communicate and express themselves, their feelings, their emotions even when they are not staying at the same place. In the absence of face-to-face contact to detect facial expressions and intonations in voice, people decipher emotions using text on online social networking sites, which have thereby emerged as a vast repository of human communication. Emotion recognition and its deep classification has recently gained much attention due to its potential applications ranging from customer satisfaction in business decision making to cognition understanding in HumanComputer-Interaction. In this paper, we present a study of different emotion detection & recognition approaches, the potential applications and the major challenges faced in this area.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Emotion Detection; Emotion Classification Models;Deep Classification; Sentiment Analysis; Text Analysis

1. Introduction Due to the profound advances in computing and communication technologies, there has been a steep growth in popularity of online social networks. Communication over online social platforms has considerably affected the way people interact with friends and acquaintances nowadays. Indeed, interacting through online social networks and online messaging systems has become ubiquitous, and a major component of a person’s life. The exponential growth of electronic communication, particularly text-based, associated with the Internet has led to a huge increase in the amount of social data repository which is not currently warehoused or mined in the ways relevant to various areas of study like Human Social Cultural Behavior analysis (Liu et al., 2012). There has been a recent swell of interest in the automatic identification and extraction of emotions, attitudes, opinions, and sentiments in textual data. Strong motivation for this task comes from the aspiration to provide tools and support for information analysts in government, commercial, and political domains, who want to be able to automatically track attitudes and feelings in the news and on-line blogs and forums. How do people feel about the recently launched Direct Benefit Transfer of LPG (DBTL) scheme in India? Is there a change in the customer satisfaction for AirAsia since its plane crashed in the Java Sea? Trying to answer these types of questions could be profoundly made easier if we have a system that could automatically detect and extract opinions, sentiments and emotions rather than sifting through the vast repository of news and data. Researchers from AI (especially Natural Language Processing and Affective Computing) have been working on the automatic detection of emotions and their classification, though it is in fact an interdisciplinary problem involving researchers from computer science, biology, psychology, cognitive science and so on.

* Corresponding author. Tel.: +91 9797 078253.

E-mail address: [email protected].

ISBN: 978-93-82288-63-3

Wani and Hashmy/COMMUNE – 2015

2. Emotion Classification Models An emotion is a certain human feeling arising out of certain mental and physiological processes in response to internal or external events that characterizes a state of mind, such as joy, anger, love, fear and so on (Agarwal et al., 2012). Classification of emotions from data first requires the choice of an emotion model for which the major consideration is the accuracy with which models represent emotions. Although the number of emotions encountered in real life can be quite large yet many theorists in this field have suggested a set of basic or fundamental emotions and are of the opinion that all other emotional states are a mixed manifestation of theses fundamental emotions. There are two considerably different models for representing emotions: the categorical model and the dimensional model. Each type of model helps to express a unique aspect of human emotion and both of them can provide an insight into how emotions are represented and interpreted within the human mind. The categorical model and dimensional models employ two different methods for estimating the actual emotional states of a person. In the categorical model, a subject is usually required to choose one emotion out of a set of emotions that best represents the feeling conveyed whileas in dimensional model employs the rating scales for each dimension by using tool like Feeltrace (Cowie et al., 2000). A brief description of different emotion models which have proved to be of particular interest for the purpose of emotion classification is presented as under. 2.1

Categorical Emotion Models

Categorical model encompasses identifying emotions by simply making the use of emotion-denoting words, or category labels. This model either assumes that there are fundamental and discrete emotional categories such as six basic emotion categories namely anger, disgust, fear, joy, sadness and surprise (Ekman, 1992), or uses expressive categories that are domain-specific. There are both primary and unrelated emotions in the model. Each emotion is uniquely distinguished by a specific set of features, expressing eliciting conditions or responses/reactions. Most work, till date, in affective computing has focused on the six basic emotions as proposed by (Ekman, 1992). However, many researchers have argued that different sets of emotions are required for different domains, for instance, in the field of teaching and education. For instance, D’Mello et al in 2007 proposed five categories (boredom, confusion, delight, flow, and frustration) for describing affect states in the student-system dialogue. Learners rarely feel sadness, fear, or disgust, whereas they typically experience boredom or delight, which is an argument for the need for domain-specific categories. The principle advantage of categorical representation model is that it represents human emotions intuitively with easy to understand emotion labels. However, it has several weaknesses due to the limited number of labels. For example, the emotional categories consist of discrete elements, and a great variety of emotions within each discrete category can be frequently observed. The categories do not cover all emotions adequately because numerous emotions are grouped together under one category. Furthermore, the same affective states can be expressed by means of dissimilar emotional categories owing to linguistic, environmental, cultural or personality differences, which leads to meager and poor agreement among emotional categories. These findings indicate that representative set of emotional categories may not characterize distinct affective states although the set of emotion categories is defined. In addition, this incomplete and hence problematic conceptualization may result in non-optimal or inefficient affect-detection. Firstly, it can potentially lead to a forced-choice identification problem, wherein the subjects are likely to discriminate among presented categories rather than to identify an emotion label themselves. This can force the subjects to choose an inappropriate and irrelevant category. The second problem is more serious and related to the first one. It is occasionally not possible for subjects to select an appropriate category since it does not exist in the label set. Therefore, a categorical model has the limitations of an identification task in attempting to identify the precise emotional states perceived by people. For instance, subjects cannot help selecting one of six basic emotions (e.g. anger, disgust, fear, joy, sadness, and surprise) even though they feel neutral and want to choose that category. Nevertheless, the categorical model with its many variations has emerged as a dominant one due to its simplicity and familiarity. 2.2

Dimensional Emotion Models

Dimensional model represents affects in a dimensional form. Emotional states in this model are related to each other by a common set of dimensions and are generally defined in a two or three-dimensional space with each emotion occupying a location in this space. A variety of dimensional models have been proposed and studied till date (Russell, 1980) (Mehrabian, 1996), (Plutchik, 1980). In Russell’s model (Figure 2.1) of affect is introduced as a reference circumflex through a figure with a setting of points representing the emotions. Emotion-related terms in Russell’s model are organized in a circumplex shape which enables a subject to choose a position anywhere between two discrete emotion-related terms. Numerical data are obtained from the relative position of the points in the two-dimensional bipolar space (valence-arousal). The valence dimension depicts positive and negative emotions on different ends of the scale while as the arousal dimension differentiates excited vs. calm states. The proximity of two emotion categories in the circumplex represents conceptual similarity of the two categories The representation and description of emotional states by means of emotion dimensions has certain advantages. A major advantage of dimensional models is that they are not correlated to a certain emotional state (e.g. angry or happy).

[426]

Wani and Hashmy/COMMUNE – 2015

Two or three dimensions of emotional meaning are commonly identified by means of rating. Due to their gradual nature, emotion dimensions are able to capture subtle emotion concepts that differ only slightly in comparison with broad emotion categories. Emotion dimensions can depict very specific identification and a large range of people’s emotion concepts. In particular, a dimensional description is well suited for the task of measuring the full-defined emotional states. Furthermore, emotional states are related to each other on a dimensional space, which is a significantly different approach from the categorical model. A dimensional model provides a means for measuring the degree of similarity between emotion categories; adjacent categories on the space are very similar, while as the opposite, categories are distinctly different from each other. Conclusively, a dimensional emotion model is a useful representation capturing all relevant potential emotions and offers a means for measuring similarity between affective states.

Fig. 2.1 Russell’s circumplex model of emotion (Russell, 1980)

Fig. 2.2 Plutchik’s emotion wheel (Plutchik)

3. Emotion Detection methods and their challenges Over the past several years, much research has been done utilizing linguistics, machine learning, information retrieval, and other theories to detect emotions. Experiments conducted so far have shown that, computers although in a coarse way, can detect and recognize emotions from texts like humans. However, all methods have certain limitations, and they lack context analysis to refine emotion categories with existing emotion models, where much work has been done to put them computationalized in the domain of believable agents. (Kao et al., 2009). Currently, three approaches dominate the emotion recognition and detection task; keyword based, learning based and hybrid based approach. These make use of features mainly selected from syntactic (e.g. n-grams, pos tags, phrase patterns) and semantic (e.g. synonym sets) data to detect emotions (Binali et al., 2010). The hybrid approach attempts to make up for the weaknesses in other two approaches (Table 3.1). Hybrid model takes advantage of the fact that the syntactic and semantic information can be highly helpful for emotion detection. Contemporary methods are lacking in in-depth semantic analysis for detecting hidden phrase patterns and more research needs to be done to identify, build, and incorporate knowledge rich linguistic resources that have a focus on recognizing and detecting emotions. The main benefit of this approach is that it can yield higher accuracy results from training a combination of emotion classifiers and supplemented by knowledge-rich linguistic information from dictionaries and thesauri. An important and direct consequence of this is that it will reduce the high cost involved in using human indexers for information retrieval tasks and minimize complexities encountered while integrating different lexical resources. Table 3. Emotion Detection Methods; Strengths and Challenges Method

Strength(s) of the method

Keyword -Based



Most intuitive, easy to implement

Learning -Based



Facilitates easy implementation of classifiers by novices who can then apply the learned model to new instances.

 Hybrid



Offset the high cost involved in using human indexers for information retrieval tasks Minimize complexities encountered while integrating different lexical resources

[427]

      

Challenges for the method Ambiguity in Keyword definitions Keyword-less sentences Conflicting keywords Determination of features (emotion indicators) The number of emotion categories that can be recognized Acquisition and processing of multi-domain corpus in supervised techniques Acquisition and processing of training data corpus when supervised learning techniques are used

Wani and Hashmy/COMMUNE – 2015

4. Conclusion In this paper, we briefly surveyed the various emotion detection models and detection and classification approaches, which are popular in the research community in this field. A comparison of different emotion detection approaches was presented. We outlined the fact that the syntactic and semantic information can be beneficial for emotion detection and classification task. References Agarwal, A., An, A., 2012. Unsupervised Emotion Detection from Text using Semantic and Syntactic Relations, In proceedings of the 2012 2012 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT). Washington, DC. Binali, H., Wu, C., & Potdar, V., 2010. Computational approaches for emotion detection in text, 4th International Conference on Digital Ecosystems and Technologies (DEST). Dubai, UAE. doi:10.1109/DEST.2010.5610650 Cowie, Roddy, Douglas-Cowie, E., Savvidou, S., McMahon, E., Sawey, M., & Schröder, M. "FEELTRACE': An instrument for recording perceived emotion in real time. In ISCA Tutorial and Research Workshop (ITRW) on Speech and Emotion. D’Mello, S., Picard, R., & Graesser, A., 2007. Towards an effect sensitive auto tutor, IEEE Intelligent systems 22, p. 53-61. Ekman, P., 1992. An argument for basic emotions, Cognition & Emotion, 6, p. 169-200. Kao, E. C., Liu, C. C., Yang, T. H., Hsieh, C. T., Soo, V. W., 2009. Towards Text-based Emotion Detection a Survey and Possible Improvements, In International Conference on Information Management and Engineering. Kaula lumpur, Malaysia. p. 70-74 doi: 10.1109/ICIME.2009.113 Liu, X., Tang, K., Hancock, J., Han, J., Song, M., Xu, R., & Pokorny, B., 2012. SocialCube: A Text Cube Framework for Analyzing Social Media Data. In proceedings of ASE International Conference on Social Informatics. Washington, DC. Mehrabian, A., 1996. Pleasure-arousal-dominance: A general framework for describing and measuring individual differences in temperament, Current Psychology 14, p. 261-292 Plutchik, R., 1980. Emotion: A psychoevolutionary synthesis, Harper & Row, New York, p. 440 Russell, J. A., 1980. A circumplex model of affect, Journal of personality and social psychology 39, p. 1161

[428]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16-18March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Achievements and Limitation of the First Machine Translation System to convert Hindi into Dogri Preeti*, Devanand Department of CS&IT, Central University of Jammu, Jammu, India

Abstract The Hindi-Dogri machine translation system has been developed to facilitate translation of Hindi text into Dogri text. A user can get the translations done on the click of a button. The development of this system is an effort to bring the Dogri language on the map of machine translation. This paper summarizes the achievements and limitations of this research work.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Dogri; Machine translation, Achievements,Limitations

1.

Introduction

The machine translation system for Hindi-Dogrihas been developed using ASP.Net and the databases are in MSAccess with Hindi and Dogri text in Unicode format. It is based on the direct approach of machine translation. It takes Hindi text as input and provides Dogri text as output in Unicode. This is the first machine translation system developed to convert Hindi text into Dogri. Both Hindi and Dogri use the Devanāgarī Script. These are closely related languages and share not only the script but also the lexicon. The major difference lies in the inflections of the words in these languages. The phases involved in Hindi-Dogri MTS are: pre-processing of the source text (Hindi), word to word translation of the text to the target language (Dogri), followed by inflectional analysis and ambiguity resolution then handling peculiarities of Dogri followed by output generation. The pre-processing phase involves text normalization, proper noun recognition and handling of collocations.(Dubey, 2014) 1.1 Pre-processing Activities 





Text Normalization: Text normalization refers to keeping a standard spelling for spelling variations of the same word. Therefore, all the variants of the same word need not be stored in the dictionary database. A database of such words has been created and has a collection of 400 words. It contains a standard word for its variants. The standard word has its meaning in the dictionary database. Some examples of such words are:गई/gī, गयी/ gayī;ह द िं ी/ hiṃdī, ह न्दी / hindīetc, Identifying Collocations:Words that cannotbe translated word to word are called collocations.They need to be handled properly for accuracy of the output. If these words are translated word to word, the word sense is changed. e.g. आपके(āp ke) should be translated into Dogri as तिं’दे / tuṃ’de. If not handled properly, it will be translated word to word asतसदे / tusa de. Proper Noun Identification: After checking for collocations in the source text, the system extracts proper nouns such as names of months, countries, days of a week, universities, banks etc. These words are usually transliterated to preserve their meaning. In our system both the source and target language use the same script, therefore these words are not to be translated; word to word so as to preserve the word sense. In this sub phase,

* Corresponding Author. Tel:+919419180084. Email address: [email protected].

ISBN: 978-93-82288-63-3

Preeti and Devanand/COMMUNE – 2015

the proper nouns are identified by referring the proper noun database. These words are not sent for further translation. The architecture of the system is presented in Fig I. 1.2 Tokenizer It segments the input stream of characters into single meaningful units called tokens. The output of the preprocessing phase goes to the tokenizer, which extracts individual words from the sentence for further processing for translation into the target language. The tokens are extracted from the text using space, a punctuation mark, as delimiter. 1.3 TargetText Generation The translation engine consists of the following sub-phases: a) Lexicon look up b) Ambiguity Resolution c) Inflectional Analysis d) Handling special cases pertaining to Dogri. i. Handling kar at the end of words ii. Handling words before raha iii. Handling words before laga The final output i.e. Dogri text is generated after going through all the above mentioned phases.The system has been tested using the human evaluation method. Both Quantitative (include intelligibility testing and accuracy testing) and Qualitative tests (include WER and SER) have been performed on the system. The accuracy of the system on the basis of Intelligibility test has been calculated as 98.54% and the accuracy test reports 98.71% accuracy. In the quantitative tests the Word Error Rate is found out to be 2.011% whereas Sentence Error Rate is 20.40% (Dubey, 2014).

Fig I: Architecture of Hindi-Dogri MTS

2.

Achievements in the Course of Development of Hindi-Dogri MTS

The achievements during the course of development of Machine Translation System for Hindi-Dogri language pair are discussed the following section:  The first survey to study the Digital Divide factors has been conducted in Jammu, to The efforts made by the Government to bridge the digital divide have also been presented (Dubey et al, 2011a).  The first Grammatical and inflectional analysis of Hindi and Dogri has been done to study the closeness between these languages (Dubey et al, 2011b ).

[430]

Preeti and Devanand/COMMUNE – 2015

 

3.

The first digital dictionary for Hindi- Dogri has been developed(Dubey, Devanand, 2013). There was no dictionary available for the language pair for Machine Translation purpose. Rules for morphological analysis of Hindi-Dogri have been devised, which can be used in future for other research work on translation of Dogri. Modules for handling Proper Nouns and collocations have been developed successfully. These modules facilitate increased efficiency of the system. Sometimes, during translation proper nouns and collocations are translated leading to inaccurate results. The module for handling proper nouns identifies proper nouns and does not send them for further translation. The module for handling collocations does the same when collocations are seen in source text. 3. Limitations of the Hindi-Dogri MTS

The system shows good accuracy of translation, but it still has some limitations. Some common errors are explained with examples: 3.1 Proper noun Recognition Failure The names that are not present in the Dogrinames database, but have their translations in the dictionary database will be translated. For identifying a proper noun, it must either be preceded with a title or be succeeded with a surname. In the absence of both (surname and title) is not identified; and hence translated. For example: Input sentence:

विजयआयाथा (vijay āyā thā)

Output Sentence:

जजत्तआएदा ा (jitt āedā hā)

3.2 Ambiguity Resolutions Words with multiple meanings are not resolved by the system. For example: Input sentence:

कैसेख्यालोंमें खोर ीथी।

Incorrect Output:

कक ’यािंख्यालें चखोआदी ी।( ki ’yāṃ khyāleṃ ca khoā dī hī )

Correct Output:

कने े ख्यालें चखोआदी ी।(kanehekhyāleṃ ca khoā dī hī)

(kaise khyāloṃ meṃ kho rahī thī)

33 Disagreement of दा/da postposition before Verb phrase: In Dogri, all the Verb phrases in the sentence must agree with the postposition दा.However, in some cases, it fails as shown in following example: Input sentence: ब नकीशादी ोगई(bahan kī śādī ho gī) Incorrect Output: भैनदीब्याह् ोईगेई(bhain dī byāh hoī geī) Correct Output: भैनदाब्याह् ोईगेआ(bhain dī byāh hoīgeā) 3.4 Gender Disagreement: The output of the system sometimes does not reflect the correct gender of a word and therefore causes gender disagreement with verb/postposition in the target language. For example: Input sentence: उन् ें रोडआइलैंडमें एकशिममला ै / (unheṃ roḍ āilaiṃḍa meṃ ek śav milā hai) Incorrect Output:उ’नेंगीरोडआइलैंडचइकलाशममलेआऐ/ (u’neṃ gī roḍ āilaiṃḍa ca ik lāśmileāai) Correct Output:उ’नेंगीरोडआइलैंडचइकलाशममलीऐ/(u’neṃ gī roḍ āilaiṃḍa ca ik lāśmilīai) 3.5 ी Inflection Missing In Dogri, most of the words end with ी . Therefore, there is a need to add this inflection in many words. Some failure cases are discussed below: Input sentence: इसमें बािंधलो। (isameṃ bāṃdh lo) Output Sentence:एह्दे चब’न्नलैओ। (ehde ca b’nn laio) [431]

Preeti and Devanand/COMMUNE – 2015

Correct output:एह्दे चब’न्नीलैओ।(ehde ca b’nnīlaio ) 3.6 Wrong Sentence Structure In Dogri, if र ा,र ी, र े are preceded with न ीिं, the structure of the sentence changes in the following manner. Input sentence: गठरीबिंधन ीिंर ीथी। (gaṭharī baṃdha nahīṃ rahī thī) Incorrect Output: गिंढब’न्ननेईंरे ी ी।(gaṃḍha b’nna neīṃ rehī hī ) Correct Output: गिंढब’न्नोआदीनेईं ी। (gaṃḍha b’nnoā dī neīṃ hī) 3.7 Disagreement of Subject Noun Phrase with Verb Phrase Some failure cases due to disagreementof all the verb phrases in the sentence with the subject noun phrases are seen in some translations. The following example illustrates the disagreement of subject noun phrase with verb phrase Input sentence: उसेक्यासजादीजानीचाह ए।(use kyā sazā dī jānī cāhie) Incorrect Output: उस्सीकेह् स’जाहदत्तीजानीचाह दा।(ussī kehs’jā dittī jānī cāhidā) Correct Output: उस्सीकेह् स’जाहदत्तीजानीचाह दी। ( ussī kehs’jā dittī jānī cāhidī) 3.8 Limited Dictionary The Hindi-Dogri dictionary used in this system consists of 18510 words, which is not very large. The efficiency of the system can be improved by increasing the size of the dictionary. 4.

Future Directions

Although this system is showing good results, using the direct translation approach but still there is lot of scope for improvement. Following are some of the future directions:  



 

5.

Increase in Database size: The present databases used by the system such as dictionary, proper noun database, surnames, titles, bigrams and trigrams for WSD can be extended to improve the accuracy of the system. Use of Newer Methods of Translation:Statistical Machine Translation approach is the latest approach of MT and is believed to give highly accurate results. The development of SMT system requires a high quality parallel corpus. Thus, with the use of the present system, parallel corpus for Hindi-Dogri Language pair can be developed for use in future researches. Use of Automatic Evaluation Techniques: Automatictechniques of evaluation such as BLUE, METEOR etc are used worldwide for evaluation of translation software.These techniques also require parallel corpus for the language pair undertaken. These techniques can be adopted in future for Hindi-Dogri Language pair, once the parallel corpus has been developed. Word Sense Disambiguation: WSD should be extended to include polysemous words also; this will increase the accuracy of the system. Though not many polysemous words were encountered during testing; but adding this module will disambiguate such words also leading to increase in translation accuracy. Website Translation:The Machine translation System can further be extended for translating websites and emails. Conclusion

This paper in brief introduces the Hindi-Dogri MTS, the achievements in the course of its development and the limitations of the system, some of which can be handled in future. Future directions of this work have also been presented. References Dubey,P., 2014. Study and development of a machine translation system from Hindi language to Dogri language: an important tool to bridge the digital divide, PhD Thesis, Department of Computer Science &IT, University of Jammu. Dubey, P., Devanand, 2013. Machine Translation System for Hindi-Dogri Language Pair, in proceedings of IEEE Conference (ICMIRA), held at SMVDU, in Dec 2013, ISBN: 978-0-7695-5013-8, pages: 422-425, DOI 10.1109/icmira.2013.89 Dubey, P., Jyoti,J., Devanand,2011a. A Study to Examine digital Divide Factors: Jammu and Kashmir Perspective, BVICAM’s International Journal of Information Technology (BIJIT) , Volume 3 ,No.2, Issue 6, July-December, 2011, ISSN: 0973-565 Dubey, P., Pathania, S.,.Devanand, 2011b. Comparative Study of Hindi and Dogri Languages with regard to Machine Translation , Language In India, Volume 11:10 October 2011,ISSN 1930-2940.

[432]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Computational Aspect of different Neuroimaging Modalities to Study the Structural and Functional Brain Networks: Analysis and Measure Based on Advanced Statistical and Classical Graph Theory Approach Vinay Shukla*, Shrawan Kumar, Dharmendra Singh Department of Computer Science & Engineering, Institute of Technology & Management Chehari, Maharajganj UP, India

Abstract Advancement in brain imaging technology gives a good insight of inner brain circuits and its functional neuronal network. Brain is a nonlinear dynamical system, considering the dynamical aspect one can also think brain as hub of neuronal networks. In recent years there has been huge study on network modelling of brain connectivity. One of the challenge to establish the network connections between heterogeneous brain regions. The purpose of this study is to explore the structural and functional brain networks using task based and non-task based data of MRI, Diffusion MRI and EEG. The temporal resolution is very promising in EEG while MRI is only give spatial resolution but combined these two imaging modality. Using EEG for functional network connectivity, has some issues and severely limited by volume conduction and its accuracy is entirely depends on source modelling. Considering the above lacuna we are trying to propose a novel approach based on interacranial cerebral EEG (SEEG) recordings in human brain. We are also delineating theoretical graph analysis of human brain networks based on various neuroimaging modalities such as structural MRI, diffusion MRI and EEG. To investigate the salient characteristics of brain regions of preferred network pathway and its connectivity hub location. We are also proposing ML (maximum likelihood) estimation between exact functional connectivity and expectation with minimization ratio between brain functional regions. Using this study one can make a possible assumption of topological organization of human brain networks and its connected network hubs. In later stage of proposed research one can also deal with various neuropsychiatric diseases and one can also see brain development and change in network properties during different stage of age.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords : EEG; Brain Network; MRI; MLE; Graph Theory; Diffusion MRI

1. Introduction Neural function, and by extension neural codes, are constrained by connectivity. Brain connectivity is really crucial to elucidating how neurons and neural networks process information (Brandes and Erlebach, 2005). The functional connectivity in human brain is spontaneous, whole process is quite debatable. Advancement in computational neuroimaging technique (CNT) gives more uncertainty and overlook brain activity. Till now most of the studies is based on external stimuli based and which is challenge for higher cognitive challenge. All these assumptions based on average statistical random analysis. In imaging technique the task based paradigm activation is kind of manipulation which clearly result of cerebral activation of circuits, and its important for performing the task. All these studies are totally based on some short of ideal base lines (Gustavo, 2011). The BOLD activity is entirely based on the signal change during performing some task and totally depends on task using this one can see the functional network during activation which leads towards default mode network (DMN) (Raichle et al, 2001). Imaging structural and functional brain connectivity has revealed the complex brain organization into large-scale networks. Such an organization not only permits the complex information segregation and integration during high cognitive processes but also determines the clinical consequences of alterations encountered in development, ageing, or neurological diseases. Recently, it has also been demonstrated that human brain networks (Sporns et al., 2004) shared topological properties with the so-called 'small-world' mathematical model, allowing a maximal efficiency with a minimal energy and wiring cost (Guye et al., *

Corresponding author. Tel.: +91 9554 776314 E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Shukla et al /COMMUNE – 2015

2008). To understand the complexity behind the human brain networks one can only understand using various kinds of neuroimaging modalities such as Structural MRI, Diffusion MRI and EEG/MEG (Horwitz, 2003). Structural MRI deal with anatomical network of different neuronal regions and its morphology. Diffusion MRI to understand and delineate the bunch of fibre connectivity in different regions in white matters tracts. Diffusion tractography and functional/effective connectivity MRI provide a better understanding of the structural and functional human brain connectivity (Guye et al., 2008). 2. State of the Art In this section we present the most prominent work relevant to our research. The aim is to understand the structural brain network followed by functional brain network using structural MRI combined with diffusion MRI followed by EEG. There are already enough research is ongoing and also significant result are also achieved. But if we combined MRI EEG and DTI we will be able to understand better brain networks from Structural to functional. Diffusion tensor imaging is very novel neuroimaging technique, using DTI one can see the white matter track and also explore the connectivity from other brain regions. One can also do fibre tractography to understand the fibre connection by each voxel. Using Probabilistic tractography one can compute the connectivity probabilities without touching the actual white matter pathways (Rajapakse et al., 2008). EEG has better temporal resolution and time dependent signal, one can easily calculate the time dependent signal. Based on time varying signal one can easily map the scalp on brain surface. One can also combined Diffusion Tensor Imaging (DTI) with EEG to understand the connectivity voxel by voxel. We are proposing novel method to understand the brain network and its connectivity through graph theory approach and measure of these networks is going to be done by advanced statistical technique. We are proposing complex graph analysis based on binary graph which is highly usable in complex networks. The Graph theoretic analysis of functional brain connectivity has helped to identify functional hubs, which are highly connected and central to information flow and integration (McIntosh and Gonzalez, 1994). Functional connectivity studies in the frequency domain have provided evidence for a fractal organization of functional brain networks (Friston, 2005). 3. Problem Statement Advancement in brain imaging modality give us very good understanding of inner brain networks. In this paper, we are going to use multimodal imaging data sets, which is purely based on particular imaging machine. Because different imaging machines has different imaging protocol. Our basic aim of study is to understand the brain in structural to functional and its connectivity using DTI. In structural brain network we would like to understand the resting state network to understand the coordinates of the brain regions. For functional network, we would like to understand the task based activity in particular brain regions. Using above we would be able to judge the development of brain networks in different age period. We would also be able to predict the networks changes in different kinds of neurological diseases. We would like to investigate some challenging research questions like: ●Brain network changes in neurological diseases and also we would like to answer different kind of connectivity structural, functional and effective connectivity or causal connectivity and its measure in certain kind of brain disorders. ●We will also explore the functional connectivity how it maps to structural connectivity. ●The parallel analysis of structural connectivity maps of the human brain and patterns of functional and effective connectivity recorded in various conditions of cognitive activation and cognitive maps. 4. Proposed Solution: General Idea To start the preliminary work we will get all the data sets provided by our collaborators. After extracting the data sets using dicom manipulation or visualization software like Mricro. Mricron will convert all the data sets in to nifti format for visualization and analysis process. For EEG data sets we will use EEG lab software and Matlab function to extract the EEG data sets using Fourier or Wavelets transform. To extracting the EEG feature extraction for time varying signal we will use wavelet transform. Wavelets transform is better to extract the signal and modularize the EEG signals. For structural MRI we used anatomical brain template like MNI brain template to match our brain data and superimpose on it. One can use (SPM, n.d) or (FSL, n.d) software for Diffusion MRI data sets. To extract the brain network point by point. For mapping the brain scalp in to network one can use theoretical graph theory approach (Bullmore et al., 2009). As we know in graph has good features to establish the complex networks. For EEG signal we will use causality modelling over time varying signals. One can also think of EEG signal is good for time and space dependent signals to map the causal signals for each node using advanced modern algebra methods like Lie Theory and its operators specially Zeeman's causality. One can also use some short of Granger causality (Sato et al., 2010) method for EEG Signals to map the entire data sets over time dependent signal into space and time. For structural MRI we will use ICA (Independent component Analysis) to localize the brain data sets into point of interest brain regions (Rowe et al., 2006). For diffusion MRI its good to use Probabilistic method for tracking the fibres and compute probabilities of [434]

Shukla et al /COMMUNE – 2015

theses fibers without touching the pathways of these fibres. Finally one can develop tools based on Matlab functions using C++ or any other programming language to make all our queries automated for possible predictions based on given query. C and C++ is good because one can easily import or convert these functions into Matlab or vice-versa using MEX compiler. 5. Conclusion The topology of structural and functional human brain networks offers a common framework to merge structural and functional imaging as well as dynamical data from electro physiology that might allow a comprehensive definition of the brain organization to understand the differences between different modality of brain networks in multimodal neuroimaging techniques. But one can only understand the differences between underline said phenomenons if and only if to dig between structural connectivity and its coherence networks. References Gustavo, D., 2011, Emerging concepts for the dynamical organization of resting-state activity in the brain. Nature Reviews.Vol. 12. Raichle, M. E. et al., 2001 A default mode of brain function. Proc. Natl Acad. Sci. USA 98, 676–682. SPM (n. d) - Statistical Parametric Mapping url: www.fil.ion.ucl.ac.uk/spm/ FSL(n.d)-url: www.fmrib.ox.ac.uk/fsl/ Sato JR et al., 2010, Analyzing the connectivity between regions of interest: an approach based on cluster Granger causality for fMRI data analysis. Neuroimage. 52(4):1444-55 Rajapakse JC et al., 2008, Probabilistic framework for brain connectivity from functional MR images. IEEE Trans Med Imaging. 27(6):825-33. Horwitz B, 2003, The elusive concept of brain connectivity. Neuroimage 19, 466-470. Sporns, O, Chialvo, D, Kaiser, M, Hilgetag, CC, 2004, Organization, development and function of complex brain networks. Trends Cogn Sci 8, 418425. Brandes, U, Erlebach, T, 2005, Network Analysis. Springer, Berlin. McIntosh, AR, Gonzalez-Lima, F., 1994 Structural equation modeling and its application to network analysis in functional brain imaging. Hum Brain Mapping 2, 2- 22. 15. Friston, KJ. 2005, Models of brain function in neuroimaging. Annu Rev Psychol 56, 57Guye M. et al., 2008, maging structural and functional connectivity: towards a unified definition of human brain organization? Curr Opin Neurol.21 (4):393-403. Bullmore E. et al., 2009, Complex brain networks: graph theoretical analysis of structural and functional systems. Nat Rev Neurosci. 10(3):186-98. Rowe DB. et al., 2006, Multivariate statistical analysis in FM.

[435]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Study of the Emergence of Sky Computing Vivek Chalotra* Department of Physics & Electronics, Baba Saheb Ambedkar Road, University of Jammu, Jammu, India.

Abstract Sky computing is truly changing the method for computing. Numerous machine resources, for example, hardware and programming modules are gathered into the resource pool which can be used by the clients by means of the web through web programs, desktops, or cell phones. It is not a new idea; it is identified with cluster computing, grid computing standard and utility computing. All these computing paradigms have really helped in the improvement of cloud computing. With the expansive scale utilization of web everywhere throughout the world, everything can be conveyed over web utilizing the idea of sky computing as a service like electrical energy, cooking gas, water etc. In this paper, we will analyze all the advances in detail, which leads to the emergence of Sky computing.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: grid computing; sky computing; cloud computing; resource balancing; cluster computing; parallel processing.

1. Introduction We have encountered an enormous change in computing from more seasoned times till today. Long ago, large machines were kept behind the glass dividers and just the experts are permitted to work on them (Lucky, 2009). Later the idea of grid computing which permits the clients to have computing on interest as per need come in existence. After that, we got such computing, which makes resource provisioning more straightforward and on need of customer. By then, finally we got the thought of distributed computing which concentrates on the provisioning and deprovisioning of calculation, storage, information services to and from the user without user being not mindful of the way that from where these resources are coming to him/her (Singh, Kirit, 2012). With the substantial scale utilization of web everywhere throughout the world, everything can be delivered over web utilizing the idea of sky computing as a service like water, cooking gas and power and so forth. The remaining part of the paper is composed as follows: Part 2 depicts the cluster-computing paradigm including its pros and cons. Part 3 depicts grid-computing paradigm including its pros and cons. Part 4 depicts sky-computing paradigm including its pros and cons. Part 5 shows the key attributes of cluster, grid, and sky computing. Finally, conclusion is displayed. 2. Cluster Computing Paradigm Cluster computing is a kind of computing in which a couple of nodes are made to run as a one large computer. The different nodes included in cluster are ordinarily associated with one another utilizing some fast LAN's (Indu et al, 2011). There are principally two reasons of deploying a cluster rather than a standalone machine, which are execution and fault resistance. An application requires high processing regarding response time, memory, and throughput particularly when we discuss real time applications. Cluster computing gives high computation by utilizing parallel programming, which is utilization of numerous processors all the while for various or a single task. An alternate reason is fault resistance which is really the capacity of a framework to work smoothly even near any flaw. As the clusters are

* Corresponding author. Tel.: +91 9419 304382. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Chalotra/COMMUNE – 2015

the copies of comparative parts, the flaw in one part just influences the cluster's energy however not its accessibility. Along these lines, clients dependably have a few components to work with even in the vicinity of fault.

Fig. 1: Cluster computing (Indu et al, 2011) Here Fig. 1 demonstrates the general idea of cluster computing as per which a few nodes merge together and are introduced as a single interface/node to the client.

2.1.

Pros and cons of Cluster Computing

Table 1.Pros and cons of Cluster Computing S.No.

1.

2.

3.

Pros Manageability: It requires a great deal of effort, expense, and cash to deal with a substantial number of segments. In any case, with cluster, vast quantities of segments are consolidated to act as a single element. In this way, administration gets to be simple. Single system Image: Again, with cluster, client simply gets the vibe that he is working with a solitary framework, but in real practice, he is working with an expansive number of parts. He requires not stressing over that segments, he just needs to deal with a single framework image. High Availability: As all the parts are imitations of one another, so if one part goes down in view of any specialized reason, then some other segment can take its place and client can keep on working with the framework.

Cons Programmability Issues: This may be the situation if the parts are distinctive regarding programming from one another, and afterward there may be issues when joining every one of them together as a single substance.

Issue in Finding Fault: Because we are managing a single element, so issue may emerge when discovering fault that which of the segment has some issue connected with it. Hard to handle by a Layman: As cluster, processing includes combining diverse or same segments together with distinctive programmability, so a non-proficient individual may think that it hard to manage.

3. Grid Computing Paradigm Grid Computing is the isolation of resources from various destinations to take care of an issue that cannot be comprehended by utilizing the processing of a single machine. It utilizes utilization of different clusters that are loosely coupled, heterogeneous and are geographically scattered. Here individual client gets access to the resources (like processors, storage and data and so forth.) on interest with practically little or no idea of the concept that where those resources are physically placed. For instance, we utilize power for running cooling systems; computers and so on through wall sockets without worried about the reality that from where that power is coming and how it is being created. It is all the more prevalently known as a collection of servers that are bound together to assault a single task. Grid computing is mainly concerned about offering, gathering, facilitating and giving services to different consumers. Despite the fact that the Grid relies upon the computer systems and communication networks of the basic web, novel programming permits clients to get to machines distributed over the network. This product is called "middleware" in light of the fact that it sits between the operating systems of the machines and the applications software that can tackle a client's specific problem. As an occurrence, the Worldwide LHC Computing Grid (WLCG) – a distributed computing framework masterminded in tiers – gives a group of in excess of 8000 physicists close continuous access to LHC information (CERN, n.d). The Grid builds on the innovation of the World Wide Web, which was invented at CERN in 1989.

[437]

Chalotra/COMMUNE – 2015

Fig. 2: Grid computing (CERN, n.d) Here Fig. 2 demonstrates the tier structure of Worldwide LHC Computing Grid (WLCG)

3.1.

Pros and cons of Grid Computing

Table 2.Pros and cons of Grid Computing

S.No. 1.

2.

3.

Pros Access to Additional Resources: Grid computing can provide access to CPU, storage and many other resources as well. Resource Balancing: Grid computing framework joins different frameworks into a single system image. Grid enabled applications needs resource balancing which is done by network by scheduling jobs on nodes that are demonstrating low usage. Reliability: Grid computing sites are topographically scattered. In case, for example, there is power or cooling bafflement at one site, then that won't impact the other site, hence high reliability will be there uncommonly.

Cons Not Stable: Grid programming and benchmarks are not steady in correlation to other computing models. Its standards are yet advancing. Fast Internet Connection Required: Gathering and amassing different resources from geologically scattered destinations require fast internet connection, which brings about high financial expense. Distinctive Administrator Domains: Sometimes political issues emerge when offering assets among diverse domains. Some extra devices are needed for having legitimate adjusting and managing among diverse environments like cfengine, opsware and so forth.

4. Sky Computing Paradigm Sky computing or Cloud computing is the new computing model, which gives huge pool of dynamical versatile and virtual resources as a service on demand. The main principle behind sky computing model is to offer computing, storage, and software as a service or as a utility. We just require web to utilize these utilities. Sky is a parallel and distributed computing framework comprising of a collection of interconnected and virtualized machines that are dynamically provisioned and exhibited as one or more bound together computing resources focused around servicelevel agreements (SLA) built through negotiation between the service provider and purchasers. Sky computing cuts the operational and capital expenses and permit the IT divisions to concentrate on vital activities as opposed to keeping the server farm running (Velte et al, 2010). It gives the service on Infrastructure level, Platform level, and Software level. It gives numerous features, for example, speed, versatility of resources, parallel processing, simply pay the utilized resources, pick an alternate technology whenever to further work, every minute of every day accessibility of services, independent location, gives reliability and security and so on. Some of the areas of cloud computing are Banking sector, Insurance sector, Space exploration and Weather forecasting. Infrastructure as a service means that hardware, software and other equipments can scale up and down dynamically. Platform as a service offers high level integrated environment to build, test and deploy custom apps. Software as a service delivers special purpose software that is remotely accessible. Sky computing research addresses the difficulties of meeting the necessities of cutting edge private, public and hybrid sky computing architectures, additionally the difficulties of permitting applications and improvement stages to exploit the profits of sky computing. The research on sky computing is still at an early stage (Reddy et al, 2011). Numerous existing issues have not been completely tended to, while new difficulties continue rising up out of industry applications. Some of the challenging research issues in sky computing are Service Level Agreements (SLA’s), Cloud Data Management & Security, Data Encryption, Migration of virtual Machines, Interoperability, Access Controls, Energy Management, Multi-tenancy, Server Consolidation, Reliability & Availability of Service, Common Cloud Standards and Platform Management.

[438]

Chalotra/COMMUNE – 2015

In the Sky deployment model, networking, platform, storage, and software infrastructure are provided as services that scale up or down depending on the demand as shown in figure 3.

Fig. 3: Sky deployment model (Reddy et al, 2011) Here Fig. 3 demonstrates sky deployment model.

Pros and cons of Sky Computing

4.1.

Table 2. Pros and cons of Sky Computing

S.No.

1.

2.

3.

Pros Resource sharing: In sky computing user can scale up and scale down the resources on need basis. It offers resources, which give services to various users. Pay-As-You-Go: Clients just need to pay only for those resources which are used by them. They can demand for more resources in the event that they require later on and they can moreover release their resources after usage. Better Hardware Management: It is simple for cloud administration supplier to deal with the hardware effectively on the grounds that all machines run the same equipment [2].

Cons Less Reliability: Cloud Computing is less solid in light of the fact that it used to impart the resources to different clients. So there is possibility to take the information of a client or information of one association may blend with the information of an alternate association. Web: The principle prerequisite for clients to utilize the administrations of distributed computing is web. Clients require high speed web connection. Inaccessibility of web would result in inaccessibility of information. Non-Interoperability: On the off chance that customer put away data in one cloud then later on he/she can't move it to an alternate cloud service supplier in light of the fact that there is non-interoperability between sky based frameworks.

5. Key attributes of cluster, grid and sky computing S.no.

Attributes

4.

Installation Place

Same physical location

5.

Possession

6.

Network and bandwidth

Single owner Low latency and high bandwidth dedicated network

7.

Privacy

Password based login

8.

Discovery

Membership services

9. 10.

Service Negotiation Client Management Resource Management Resource Allocation

Limited Centralized

Grid computing High-end Servers (rack/tower/blade) 1000 computers Mainly Unix/Linux, Middleware Different physical locations Multiple owners High latency and low bandwidth dedicated network Public/Private key pair based authentication Centralized indexing and decentralized data services Yes, SLA based VO based, Decentralized

1.

Composition

Standard PCs

2.

Magnitude Operating System and software

100 computers

Centralized

Distributed

/Distributed/ Centralized

Centralized

Decentralized

Centralized/Decentralized

3.

11. 12.

Cluster computing

Windows/Linux/Unix

[439]

Sky computing Standard PCs, servers and NAS 100 to 1000 computers Multiple system software’s running on Virtual Machines Different physical location Single owner Low latency and high bandwidth dedicated network Each Application/ User is provided with a Virtual machine Membership services Yes, SLA based Centralized or third-party

Chalotra/COMMUNE – 2015

S.no.

Attributes

Cluster computing

13.

Standards

VIA based

14. 15.

Single System Image Capability

16.

Failover Management

Yes Stable Failed tasks are generally restarted

Grid computing Open Grid forum standards No Varies but high Failed tasks are generally restarted

17.

Pricing

Non-open market

Mainly internal pricing

18.

Potential

Limited

19.

Internetworking

20.

Applications

Multiple clusters within an Organization Business, Science, and Data centers

Limited, mainly scientific computing oriented Multiple grid sites composed of clusters Collaborative Scientific and HPC Applications

Sky computing Web services (SOAP/ REST) Yes, but optional On demand provision Strong support for failover and Content replication. Pricing according to service provider and client Highly potential, offer individual or combined cloud services to users Loose coupling of services from different clouds Web applications and Content delivery

Table 2. Key attributes of cluster, grid and Sky Computing [2]

6. Conclusion Sky computing is the latest innovation of computer network system, giving the web services at lower expense contrasting with typical strategy. It helps enhance the services in other related technologies, for example, Grid, cluster and utility computing. In this paper, we highlighted the pros, cons and looked at the various attributes of cluster computing, grid computing, and sky computing.

References Lucky, R. W., 2009. Reflections Cloud computing, May 2009, IEEE Spectrum. Indu, G., Pawanesh, A., Pooja, G., Rohit, U. and Sandeep, S., 2011. Cloud Computing Over Cluster, Grid Computing: a Comparative Analysis, Journal of Grid and Distributed Computing, pp-01-04. CERN, Conseil Européen pour la Recherche Nucléaire, url:http://home.web.cern.ch/about/computing Velte, A.T., Velte, T.J. and Elscnpeter, R., 2010. Cloud Computing- A Practical Approach, The McGraw-Hill Companies, New York. Singh J. Y., and Kirit, M., 2012. Cloud Computing Concepts, Architecture and Challenges, International Conference on Computing, Electronics and Electrical Technologies [ICCEET], IEEE. Reddy, V. K., Rao, B. T., Reddy, L. S. S. and Saikiran, P., 2011. Research Issues in Cloud Computing, Global Journal of Computer Science and Technology, Volume 11, Issue 11.

[440]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Digital Identity Called Selfie-Means of Narcissism, Self-Exploration or Entertainment? A Review Aadil Masood Wani, Benish Ali Bhat* MERC, University of Kashmir, Srinagar, India

Abstract Digital technology has converted Selfie into a phenomenon. This paper seeks to trace the dual nature of a Selfie which simultaneously can be a social discourse and a profound personal communication tool. A Selfie has a unique nature as it can defy time and space by being narrowcast and broadcast at the same time which gives an individual a power to share and portray the ‘self’. This paper firstly looks at the identity created by a Selfie and whether it has the power to bring out egotism in personalities. Secondly, many people tend to look at selfies as a medium of documenting and exploring individuality and the paper gazes into that aspect. Finally, the paper touches the entertainment aspect of the Selfie where celebrities use it as a tool to casually connect with their audiences and they in turn are offered an opportunity by the Selfie to grab their slice of fame.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Selfie; Narcissism; Self Exploration; Online Social Networks

1. Introduction Certain tilt in the front facing cell phone camera, usually about 45 degrees above the eyeline, (Day, 2013) a light source, a good pose or may be a pout, a click, a flattering filter and a new Selfie is ready to go on your social network profiles. Facebook likes and the Instagram hearts are the aim of this new Selfie. According to Berit Skog, 2013, receiving many likes on Facebook can be interpreted as the amount of appreciation gotten from Facebook friends. It subsequently develops an identity and appreciation of the self. Communication solely through the written word or texts is long gone and ‘photographs are dominating all dialogues’ now. (Fuerst, 2013) "Selfie" was named as word of the year in 2013 by Oxford Dictionaries. "Language research conducted by Oxford Dictionaries editors reveals that the frequency of the word Selfie in the English language has increased by 17,000% since this time last year," Oxford wrote in justifying its choice. The choice was definitive and raised a dialogue throughout the world about the importance and the consequent formation of the digital identity by the Selfie. The same dictionary defines a Selfie as, “A photograph that one has taken of oneself, typically with a smartphone or webcam and uploaded to a social media website”. The definition clearly indicates to the technologies that have made Selfies rampant i.e. smartphone or a webcam and social media website. Self-portraits are not a new phenomenon but owing to the growing technology camera id no more a luxury. Robert Cornelius, a ‘Ductch born US chemist’ made the ‘first self-portrait in 1839’ using the ‘daguerreian light process’ (Cahill, 2013). Then came the Polaroid cameras that ‘freed individuals from the photographic dark rooms’ adding ease but the boom of ‘self-portraits came with the invention of the compact digital cameras’ and then came the Selfie revolution when ‘Apple launched their iphone4 with a front facing camera’ (Day, 2013) Smartphones grew and so grew the amalgamation of different technologies. The camera came together in these devices with the vast opportunities of sharing the photographs across social networks. In short people got an access to a platform where they could turn there their daily activities and their looks into viewable commodities and create their own digital identities. R. Swaminathan puts the process precisely as:

* Corresponding author. Tel.: +91 9796 722980. E-mail address:[email protected]. ISBN: 978-93-82288-63-3

Wani and Bhat/ COMMUNE-2015

“What animates a selfie from a mere picture to a complex life world of multi-vocal meanings is its unique ecosystem of production, distribution, and consumption. The camera-embedded smartphone is its means of production, mobile Internet-enabled social media is its distribution platform and its consumption is through a variety of networked smart devices. This unique digital DNA of the selfie, analogous to a human genetic code, allows it to exist in multiple forms and spaces simultaneously” 2. Selfie and Narcissism The lookout for self-worth and admiration is a part of human nature but as it slips out of hand, narcissism comes into play. ‘Narcissus a beautiful Greek youth’ fell in love with his own reflection and pined away ‘yearning for a mere image’ and so the word narcissism has its roots in there and it is ‘recognized as a psychiatric personality disorder’ by the American Psychiatric Association (Race, 2002). Given the rise of social networking sites and the smartphones with cameras that make them accessible at all times and all places, ‘Narcissists need not wait’ until others are available to ‘engage in self-aggrandizement, but can instead curate, manage, and promote an online “self” throughout the day’ by posting selfies (Panek, Konrath, 2013). According to Mark R. Leary, Professor of Psychology and Neuroscience at Duke University and author of The Curse of the Self: Self-Awareness, Egotism, and the Quality of Human Life i “By posting selfies, people can keep themselves in other people’s minds. In addition, like all photographs that are posted on line, selfies are used to convey a particular impression of oneself. Through the clothes one wears, one’s expression, staging of the physical setting and the style of the photo, people can convey a particular public image of themselves, presumably one that they think will garner social rewards.” The realisation of self-worth through the number of likes or comments you get is common among the youth today. And across the various platforms like Facebook, Instagram and Snapchat etc. the kind of selfies posted differ. According to José van Dijckii , Professor of Comparative Media Studies at the University of Amsterdam, Facebook features mostly normal and ‘ordinary self-portraitures’ while as Instagram is for “stylish selfies or stylies” and Snapchat selfies are more like ‘funny postcards’. Taking ‘excessive selfies and posting’ them has also been seen as a ‘psychiatric disorder’ (Graham, 2014). In an Article written by Graham for Mail Online, Dr David Veale, a consultant psychiatrist at the South London and Maudsley NHS Trust and The Priory Hospital, points out that people taking excessive selfies suffer from Body Dysmorphic Disorder (BDD)iii and says, “Two out of three of all the patients who come to see me with BDD since the rise of camera phones have a compulsion to repeatedly take and post selfies on social media sites.’ In the same article Graham refers to a British teenager Danny Bowman who tried to ‘commit suicide because he was unsatisfied with his appearance in the selfies he took’. According to recent findings from the Pew Research Centre, ‘teenagers in America are sharing more information than ever’ about themselves on social media (BBC News, 2013) and of those studied, 91% post photos of themselves online - up from 79% in 2006. Since narcissistic disturbance involves an ‘intense need to gain recognition and admiration through some form of exhibiting one’s self’, Selfie coupled with social media allows for an endless opportunity to gain both the ‘superficial attention that a narcissistic person may crave’, as well as an ‘easy avenue for manipulating one’s image’ (Beck, 2014). Seeking ‘validation is normal’ but with social media it can easily ‘spiral out of control’ and become a narcissistic addiction says Dr Jill Weber a psychologistiv . Getting liked on Facebook or Instagram seems to fill the individual with ‘reassurance and approbation’ which is very ‘addictive’ and the ‘cycle is repeated’ (Day, 2013). 3. Selfie as Means of Self-Exploration Not all find Selfies useless means of show-off but certain scholars find in them an ability to give expression and to provide a better perspective to study the human race. For the future generations the selfies might be ‘important historical documents’ to understand the current era (Cahill, 2013). The view that Selfie is nothing more than an ‘outlet for self-expression’, which just happens to be ‘shared more publicly via the communication modes of our times’, is catching up (Sifferlin, 2013). “Self-captured images allow i Oxford University Press blog features the scholarly reflections about their choice of the word of the year. Retrieved from: http://blog.oup.com/2013/11/scholarly-reflections-on-the-selfie-woty-2013/ ii ibid iii BDD is characterised by a preoccupation with one or more perceived flaws in appearance, which are unnoticeable to others, according to the BDD Foundation. As well as the excessive self-consciousness, individuals with BDD often feel defined by their flaw. They often experience an image of their perceived defect associated with memories, emotions and bodily sensations – as if seeing the flaw through the eyes of an onlooker, even though what they ‘see’ may be very different to their appearance observed by others. Sufferers tend repeatedly to check on how bad their flaw is - for example in mirrors and reflective surfaces - attempt to camouflage or alter the perceived defect and avoid public or social situations or triggers that increase distress. iv Taken from an article by Melissa Walker, ‘The Good, the Bad, and the Unexpected Consequences of Selfie Obsession’ written for ‘Teen Vogue’ in August 2013

[442]

Wani and Bhat/COMMUNE – 2015

young adults and teens to express their mood states and share important experiences,” says Dr. Andrea Letamendi, a clinical psychologist and research fellow at UCLA. She adds that developmentally, selfies make sense for children and teens. And for the most part, they are ‘simply reflections of their self-exploration and nothing more’ (Sifferlin, 2013). According to Dr. Rutledge, 2013: “One of the most effective ways to know yourself is to see yourself as others see you. Selfies offer the opportunities to show facets of yourself, such as the arty side, the silly side, or the glamorous side. We learn about people by accumulating information over time. Our understanding of everything, include other people, is a synthesis of all the things we know about them. By offering different aspects through images, we are sharing more of ourselves, becoming more authentic and transparent—things that digital connectivity encourages.” Experts also find that selfies can be used to look into the psychological state of individuals especially ‘adolescents’ and provide a ‘deeper window into their issues’ (Sifferlin, 2013). Selfies are not always about self-absorption, they can easily be ‘artistic expressions’ or ‘fashion statements’ (Rutledge, 2013). By sharing selfies individuals can also be expressing the desire to be from a ‘certain community or a group’ (Tifentale, 2014). Fink, 2014 says that the selfie is a form of documentation, a modern diary and there is nothing inherently wrong or negative about documenting one’s life. This way a selfie can be a catalyst for introspection. A more appropriate reaction to this new form of documentation according to Fink is to attempt to maximize it and use every selfie as an opportunity for self-reflection. 4. Selfie and Entertainment: In Edelman'sv eighth annual study on how and why people consume and share entertainment, it was found consumers in the U.S., UK and China want their entertainment "selfie-style"-- centered on the individual, immediately gratifying, engaging and sharable across social networks (Becker, 2014). Celebrities have always taken to social media to attract their audiences and they are equally fond of Selfies as a medium to reach out. Celebrity Twitter and Instagram profiles are common and filled with their selfies. International celebrities like Rihanna, Justin Bieber, Lady Gaga and Madonna are all ‘serial uploaders’ of selfies (BBC News, 2013). In their ‘unique style’ and ‘casual rawness’, selfies feel more ‘immediate’, ‘intimate and personal’, enhancing the ‘celebrity’s connection’ to their fans (Rutledge, 2013). The famous TV celebrity Kim Kardashian is releasing a 352-page book of curated selfies called Selfishvi . According to R. Swaminathan, Kardashian may not have intended it as such, but the title of the book is richly infused with interpretative possibilities for the fields of social and philosophical anthropology and the complex notions of manufactured self and selfhood. As a relational social construction the selfie is a product of popular culture: the song #selfie by Chainsmokers7 reached the Billboard Top 10vii . It’s also a material, non-material and ‘epistemological foundation’ for creating ‘a sense of meaning’ of daily life and human interactions (Swaminathan, n.d.). As Becker, 2014 puts it in her summary of the Edelmans report, Today's global consumers expect unprecedented control over what they watch and when and where they watch it. They want content that is instantaneous, self-revolving, engaging in the moment and engaging others at their choosing. In short, they want entertainment "selfie-style." 5. Conclusion The selfie is increasingly becoming a symbol of a slightly ‘shifting sense of self’, one that is more aware of how we always function in at least two modes at once, the private and the public, the internal and the external (Alang, 2013). Looking at all the three aspects of the Selfie it can be concluded that its pervasive nature is because of the availability of smartphones and social media and it has changed the way we conceived communication. Whether as narcissism or expression or entertainment Selfie has become an important part of the digital footprint and has given rise to alternate digital identities. The people living in the two worlds- the real and the virtual are maintaining the two images constantly; the selfimage and the selfie-image. Technology aspires to improve human life but there is a thin line between the good it brings and its negatives. An autobiography of the ‘self’ through images and not just words is a powerful tool that technology offers. However, the constant dependence on how others value ‘you’ through their likes and comments destroy the very ‘self’ that one wants to explore. This is where the thin line between self-exploration and narcissism blurs. Albert Einstein quotes, “I fear the day technology will surpass our human interaction. The world will have a generation of idiots.” v

Edelman is an international public relations firm founded and named after Daniel Edelman and currently run by his son Richard Edelman. www.latimes.com/entertainment/gossip/la-et-mg-kim-kardashian-selfie-book-20140808-story.html vii Please watch: http://www.youtube.com/watch?v=kdemFfbS5H0 Accessed on January 05, 2015 vi

[443]

Wani and Bhat/ COMMUNE-2015

Is the Selfie entertainment and self-exploration required or are we actually breeding generations of narcissistic Selfie idiots? The link between Selfie and growing self-love leading to narcissism is an area that needs a proper thought and research. The fact that Selfie was word of the year in 2013 shows the power of an image and subsequently leads us to the understanding that human nature seeks validation and pleasure. The heights of it give it a narcissistic form, moderate use make its self-exploration and the fact that it is just pleasure makes it pure entertainment. References Self-portraits and social media: The rise of the 'selfie'. (June 7 2013). Retrieved January 14, 2014, from BBC News: http://www.bbc.co.uk/news/magazine-22511650 How to Spot a Narcissist Online. (January 16 2014). Retrieved January 17, 2014, from The Atlantic: http://www.theatlantic.com/health/archive/2014/01/how-to-spot-a-narcissist-online/283099/ ALANG, N. (November 26 2013). You are wrong about 'selfies,' they are not proof of narcissism. Retrieved January 17, 2014, from theglobeandmail: http://www.theglobeandmail.com/technology/digital-culture/you-are-wrong-about-selfies-they-are-not-proof-of-narcissism/article15600483/ Beck, J. (January 16 2014). How to Spot a Narcissist Online. Retrieved January 17, 2014, from The Atlantic: http://www.theatlantic.com/health/archive/2014/01/how-to-spot-a-narcissist-online/283099/ Becker, G. (June 19 2014). Entertainment in the Era of the Selfie. The Huffington Post. Blaine, L. (August 14 2013). How Selfies Are Ruining Your Relationships. Retrieved January 17, 2014, from Time NewsFeed: http://newsfeed.time.com/2013/08/14/how-selfies-are-ruining-your-relationships/ Cahill, D. (October 17 2013). No filter needed: The origin of the Selfie. New York Post. Retrieved january 3, 2015, from http://nypost.com/2013/10/17/the-art-of-taking-selfies-is-nothing-new/ Connell, J. O. (December 11 2013). Selfie, word of 2013, sums up our age of narcissism. Retrieved January 17, 2014, from Irishtimes: http://www.irishtimes.com/life-and-style/selfie-word-of-2013-sums-up-our-age-of-narcissism-1.1623385 Cross, T. (November 8 2013). The Culture of Now – The rise of imagery in social media. Retrieved January 15, 2014, from The Wall: http://wallblog.co.uk/2013/11/08/the-culture-of-now-the-rise-of-imagery-in-social-media/ Day, E. (July 14 2013). How selfies became a global phenomenon. The Guardian. Fink, R. E. (January 5 2014). Is the selfie narcissism at its finest? Retrieved Janauary 17, 2014, from HAARETZ: http://www.haaretz.com/jewishworld/rabbis-round-table/.premium-1.567136 Fuerst, E. (2013). Cultural Hybridity: Remix and Dialogic Culture. Retrieved January 12, 2014, from blogs.commons.georgetown.edu: https://blogs.commons.georgetown.edu/cctp-725-fall2013/2013/11/04/social-medias-impact-on-photography/ Gervais, S. J. (January 22 2013). Does Instagram Promote Positive Body Image. Retrieved January 17, 2014, from Psychology Today: http://www.psychologytoday.com/blog/power-and-prejudice/201301/does-instagram-promote-positive-body-image Graham, S. (April 10 2014). Take a lot of selfies? Then you may be MENTALLY ILL: Two thirds of patients with body image disorders obsessively take photos of themselves. Retrieved from MailOnline: http://www.dailymail.co.uk/sciencetech/article-2601606/Take-lot-selfies-ThenMENTALLY-ILL-Two-thirds-patients-body-image-disorders-obsessively-photos-themselves.html Kaufman, M. T. (February 24 2003). Robert K. Merton, Versatile Sociologist and Father of the Focus Group, Dies at 92. The New York Times. Kleinman, Z. (August 16 2010). How the internet is changing language. BBC News. Knox, S. (October 30 2013). Enough with the selfies already. Retrieved January 18, 2014, from live4: http://www.live4.com.au/enough-with-theselfies-already/ LOSSE, K. (June 5 2013). THE RETURN OF THE SELFIE. Retrieved January 5, 2014, from New Yorker: http://www.newyorker.com/online/blogs/elements/2013/06/the-return-of-the-selfie.html McIntosh, A. (September 15 2013). NARCISSISTS LIKE SOCIAL MEDIA. Retrieved January 17, 2014, from murdochindependent: http://www.murdochindependent.com.au/narcissists-like-social-media/ Studies in world Chritianity. (n.d.). Retrieved from Project MUSE:https://muse.jhu.edu/journals/studies_in_world_christianity/summary/v014/14.2.malhone.html NEWS, B. (June 7 2013). Self-portraits and social media: The rise of the 'selfie'. BBC News Magazine. Nicola Bruno, M. B. (Febuary 6 2013). Self-Portraits: Smartphones Reveal a Side Bias in Non-Artists. Retrieved January 18, 2014, from plosone: http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0055141 Panek, N. K. (2013). HOW NARCISSISM DIFFERS ON FACEBOOK AND TWITTER. Retrieved January 17, 2014, from sarakonrath: http://www.sarakonrath.com/media/publications/narcissism_SNS_sites.pdf Panek, N., & Konrath, S. H. (2013). HOW NARCISSISM DIFFERS ON FACEBOOK AND TWITTER. Retrieved January 15, 2104, from Sarakonrath: http://www.sarakonrath.com/media/publications/narcissism_SNS_sites.pdf Pearse, D. (March 17 2012). Facebook's 'dark side': study finds link to socially aggressive narcissism. Retrieved January 17, 2014, from Guardian: http://www.theguardian.com/technology/2012/mar/17/facebook-dark-side-study-aggressive-narcissism Race, T. (July 29 2002). New Economy; Like Narcissus, executives are smitten, and undone, by their own images. Retrieved january 15, 2014, from New Yor Times: http://www.nytimes.com/2002/07/29/business/new-economy-like-narcissus-executives-are-smitten-and-undone-by-their-ownimages.html?src=pm Rawlings, K. (Novemnber 21 2013). Selfies and the history of self-portrait photography. Retrieved January 14, 2014, from OUP blog: http://blog.oup.com/2013/11/selfies-history-self-portrait-photography/ Rutledge, P. (April 18 2013). #Selfies: Narcissism or Self-Exploration? Retrieved January 17, 2014, from Psychology Today: Here to Help: http://www.psychologytoday.com/blog/positively-media/201304/selfies-narcissism-or-self-exploration Sifferlin, A. (September 6 2013). Why Selfies Matter. Retrieved from http://healthland.time.com/2013/09/06/why-selfies-matter/ Skog, B. (February 27 2013). What's the thing about "Like!" on Facebook. Retrieved from Popularsocialscience.com: http://www.popularsocialscience.com/2013/02/27/whats-the-thing-about-like-on-facebook/ Slavin, L. (January 17 2014). The Evolution of Selfie Culture: Self-Expression, Narcissism, or Objectification? Retrieved January 18, 2014, from feminspire: http://feminspire.com/the-evolution-of-selfie-culture-self-expression-narcissism-or-objectification/ Swaminathan, R. (n.d.). Self, Selfhood and a Selfie: The Anatomy of a Virtual Body and Digital Identity. Tifentale, A. (February 2014). The Selfie: Making sense of the “Masturbation of Self-Image” and the "Virtual Mini-Me". Retrieved from Selfiecity.net: http://d25rsf93iwlmgu.cloudfront.net/downloads/Tifentale_Alise_Selfiecity.pdf Titlow, J. P. (January 31 2013). #Me: Instagram Narcissism and The Scourge Of The Selfie. Retrieved January 17, 2014, from REadWrite: http://readwrite.com/2013/01/31/instagram-selfies-narcissism#awesm=~otc2NlnjHXOpwC

[444]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Experimental Study of Different Wavelets for Real Time Environmental Monitoring in Wireless Visual Sensor Networks Umar Farooq, Shabir Ahmad Sofi*, Roohie Naaz Department of Information Technology, NIT Srinagar, India

Abstract Wireless Sensor Networking is a modern information gathering technology that has a wide range of applications in environmental monitoring. However due to the fact that sensor nodes have limited battery and are deployed in remote and harsh environments, energy efficient and real time transmission of the information are still open challenges. Since in a wireless sensor network data transmission is the most power consuming task (80% approx.), so far most useful techniques for the purpose of energy efficient and real time transmissions are based on data compression, the majority of them are based on wavelet transform. In this paper we describe the robust use of wavelet transform and make the Performance analysis of different wavelets for energy efficient and real time image data transmission in environment monitoring using wireless visual sensor networks. Our performance evaluation shows that compared to other wavelets, Haar wavelet is better in terms of energy efficiency and transmission time.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Wireless Visual Sensor Network ; Discrete Wavelet Transform ; Skipped High-pass Sub-band ; Image Compression.

Introduction Wireless Visual sensor Network (WVSN) is a self-organized intelligent system and consists of a number of visual sensor nodes (VSNs) deployed over large environmental area that can be used to monitor environment ( Akyildiz et al 2002)(Guang-yu et al, 1999). These nodes are capable of sensing, processing, and transmitting the gathered information. Fig.1 shows the Wireless Sensor Network architecture that can be used for environmental monitoring. Sensor nodes will communicate with each other and transmit the processed data to sink node over a wireless communication link. Sink node collects data from all the nodes, and transmits the analyzed data to user via Internet (Mhatre and .Rosenberg, 2004). WVSNs are unique and more challenging as compared to other sensor networks as they produce two-dimensional data whereas in other sensor networks, the scalar data is produced. Due to large amount of data, WVSNs place more stringent requirements on power bandwidth and processing capabilities. Under the context of environmental monitoring, WVSNs are to be deployed in remote areas therefore changing battery becomes impractical and assiduous task. Hence, the energy consumption of the sensor nodes is proving to be a critical constraint in employing WVSNs. Energy consumption operation of a wireless sensor node comprises of sensing, processing, and transmission. Among these operations, data transmission is the most power consuming task (80% approx.) and it is widely accepted that the energy consumed in one bit of data transfer can be used to perform a large number of arithmetic operations in the sensor processor (Pottie and. Kaiser, 2000)( Anastasi, et al, 2009). Moreover, in a densely deployed sensor network, the physical environment would produce very similar data in near-by sensor nodes and transmitting such data is more or less redundant. Under these energy constraint conditions, it is useful to transmit only a limited amount of data bits by compressing the data using the different compression techniques like Discrete Cosine Transform, Wavelet Transform in order to reduce the energy consumption( Farooq, 2014; VijendraBabu, et al , 2008).In this context, image (2-D signal) transmission over WSN has been done mainly after using the compression algorithm based on Wavelet transform in order to reduce number of bits used to represent the image by eliminating the various redundancies, thereby reducing the energy consumption in image transfer. Adaptive compression based congestion control technique (JH and IB 2010) is one of the approaches utilizing the wavelet transform to reduce the *

Corresponding author. Tel.: +91 9419 009971. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Farooq et al /COMMUNE – 2015

number of packets so as to provide efficient energy consumption mechanism. In work (Lecuire et al, 2007) open loop and closed loop image transmission schemes based on Discrete Wavelet transform (DWT) were proposed which provide graceful trade-off between image quality and energy utilisation to transmit the image data. Skipped High-pass Sub-band (SHPS) technique (Nasri, et al, 2010) is one of the effective approaches of image transmission utilising DWT. A distributed algorithm based on lifting scheme (Ciancio and Ortega, 2004) to decorrelate the collected data at nodes by the exchange of data among other sensor nodes in network path is also an efficient approach of reducing overall energy consumption using the right trade off among local processing and transmission operations.

Fig.1 Architecture of Wireless Visual sensor Network

2. Design of Wireless Visual Sensor Network 2.1

System Overview

The system overview of the WVSN for monitoring the environment is shown in Fig.2.The architecture consists of three parts: (i) Data monitoring nodes in the sub region. (ii) Base station and (iii) Data monitoring centre. The sensor nodes are deployed in remote environment area where they collect the information of that environment in the form of image data. This image data after suitable processing techniques is sent to the monitoring centre. In each sub region, wireless network based on ZigBee technology ( Willig , 2005)is developed. ZigBee technology is preferred because it consumes significantly less power and is intended to be simpler and less expensive which makes it more suitable for wireless sensor network applications. The working frequency bands of ZigBee include 868 MHz (for Europe), 915 MHz (for USA), and 2.4 GHz (Global) ISM frequency bands.

Fig. 2 System Overview of WVSN for environment monitoring.

[446]

Farooq et al /COMMUNE – 2015

In each sub region, base stations based on the ZigBee gateway and CDMA transmission channel are deployed to perform the data collection and condition monitoring of sensor nodes in ZigBee network and transmit the parameters of the sub region to monitoring centre by CDMA wireless network. 2.2

Design of Data Monitoring Node

The system design of a data-monitoring node is shown in Fig. 3. The design consists of several units/modules such as data sensing unit, processing unit, radio frequency unit and power supply unit. The sensing unit consists of the sensors and an Analog to Digital Convertor (ADC) for data acquisition. Type of the sensors to be used depends on the application of the sensor network e.g. in case of environment monitoring the sensors required are temperature sensor, humidity sensor, Pressure sensor etc. Processing unit includes a microcontroller and memory for processing and storing of data. Power module consists of tiny battery to provide the necessary energy for other units of the monitoring node. Each node is connected with and controlled by the ZigBee communication protocol (Willig, 2005)

Fig.3 A typical data monitoring node

3. Wavelet Analysis Wavelet analysis has emerged as an important milestone in the field of spectral analysis because of its multiresolution and localisation capabilities in time as well as frequency domain. In contrast to a wave, wavelets are localized waves and have their energy concentrated in time or space and are suited to analysis of transient signals (Mallat, 1999 )( Mathieu and Daubechies,1992). The wavelet decomposition of the signal at various frequencies reveals the low and high frequency components thereby helps in localising these features in time domain. There are number of algorithms that are designed to perform such decomposition and the selection of particular algorithm depends on the application at hand. For example in case of processing of medical images or seismic signals, Continuous Wavelet Transform (CWT) is used which calculates the wavelet transform as given by the equation 3.1, where x(t) is the signal to be analyzed. Ψ(t) is the mother wavelet or the basis function. XWT (τ,s) =

𝟏 √|𝒔|

∫ 𝒙(𝒕) · Ψ*(

𝒕− 𝝉 𝒔

)dt

(3.1)

All the wavelet functions used in the transformation are derived from the mother wavelet through translation (shifting) and scaling (dilation or compression).The mother wavelet used to generate all the basis functions is designed based on some desired characteristics associated with that function. The translation parameter τ relates to the location of the wavelet function as it is shifted through the signal and hence corresponds to the time information in the Wavelet Transform. The scale parameter ‘s’ corresponds to frequency information of the transform. Discrete Wavelet Transform (DWT) is an alternate to the CWT for certain applications like image and signal compression and is usually preferred over other transforms. Although it is not the only possible choice, Discrete Wavelet transform (DWT) coefficients are usually sampled from the Continuous Wavelet Transform (CWT). There are a number of basis functions that can act as the mother wavelet for Wavelet Transformation. Since the mother wavelet produces all wavelet functions used in the transformation, it determines the overall characteristics of the resulting Wavelet Transform. Based on the number of coefficients and vanishing moments each family of the wavelets are having wavelet subclasses. Haar wavelet is the oldest and simplest wavelet. The number of vanishing moments in this wavelet is one(1). Daubechies wavelets are also called Maxflat wavelets as their frequency responses have maximum flatness at

[447]

Farooq et al /COMMUNE – 2015

frequencies 0 and π. These wavelets are completely supported with extreme phase. Sym wavelet (Symlet) is both orthogonal and biorthogonal and have compact support and has highest number of vanishing moments for given support width. Table I illustrates the characteristics of some of the wavelets which are investigated in this paper. Each of the wavelet is having a different filter length which is an important property in determining the computational cost. Obviously the wavelet filter length is directly proportional to the computational cost (Abbate et al, 1995). Table I Characteristics of wavelet functions where N is the order

3.1

Wavelet

Support width

Filter length

Orthoganality

Haar

1

2

Yes

Daubechies

2N-1

2N

Yes

Coiflet

6N-1

6N

Yes

Bior

2Nd+1 for decomp. 2Nr+1 for reconst.

Max(2Nr,2Nd) +2

No

Algorithm implementation using Wavelet Transform.

The algorithm of data transmission consists of two modules viz. encoder and decoder module as shown in Fig. 4

(a)

(b)

Fig. 4 (a) Encoder module (b) Decoder module

Image is transformed into matrix form so as to make image suitable for compression before applying the Discrete Wavelet Transform (DWT) which separates the data signal into fine-scale information known as detail coefficients and rough-scale information known as approximate coefficients ( Abbate, 1995). Since image is typically a two dimensional signal, a 2-D equivalent of the DWT is performed. This is achieved by first applying the Low Pass (L) and High Pass (H) filters to the lines of samples, row-by-row and then re-filtering the columns of the output by the same filters. As a result, the image is divided into 4 sub-bands, LL, LH, HL and HH. The LL sub-band contains the low-pass information and the others contain high-pass information of horizontal, vertical and diagonal orientation. The LL sub-band provides a half-sized version of the input image which can be transformed again to have more levels of resolution. After applying DWT, next step is to perform SHPS technique. In SHPS, attempts to conserve energy are made by skipping the least significant sub band Hi in each transform level. The low pass sub bands are further decomposed leading to LLi and LHi sub bands. By skipping two out of every four sub bands, SHPS technique reduces computational loads and data to be transmitted because only LLi and LHi sub bands are computed. The next step is quantification of the sub-bands which is used to reduce the number of bits needed to represent the image. The next step is Coding of the quantised coefficients. We use here Huffman Coding which is one of the simplest compression techniques .The main idea behind the use of this coding technique is that on one hand it gives better results in terms of compression ratio and the average bits per character than run length Encoding, Shanon-Fano coding and dictionary based techniques like Lempel Ziv Scheme (LZ77 and LZ78) (Shanmugasundaram and Lourdusamy, 2011). On other hand experimental results have revealed that although the compression ratio of the arithmetic coding for different image sizes is higher than the Huffman

[448]

Farooq et al /COMMUNE – 2015

coding but comparatively less execution time is taken by Huffman coding ( Shahbahrami et al , 2011). Finally data packets are transmitted over the Wireless Sensor network. At the decoder end, after receiving the bit stream, the first step is Inverse Huffman decoding and Finally Inverse Discrete Wavelet Transform (IDWT) is applied to these decoded coefficients to recover the image. 4. Performance Analysis In this section we implement and test the encoder and decoder module for the sensor node based on Imote2 platform for different gray scale images in MATLAB. We evaluate the energy consumption and transmission time for different wavelets using the same transmission scenario (Fig.5). Encoding and decoding is performed respectively at the source and destination node.

(a)

(b)

Fig.5(a) Original images (b) Transmission Scenario

4.1

Energy Analysis

For the energy analysis of the transmission system for different wavelets, we consider gray scale Lena image of size 128×128 with the distance between the source and destination node as 10m. The results are shown in the Table II which shows that Haar wavelet is the best option for energy efficient data transmission as the minimum energy is consumed in that case. Table II Comparative analysis of energy consumption for different wavelets Wavelet

Compression ratio 4.21 4.0 3.65 3.07 3.76 2.91

Haar db1 db4 db10 coif1 coif4

4.2

Energy(mJ)

Wavelet

Compression ratio

Energy(mJ)

19.54 20.57 22.55 26.76 21.88 28.25

Sym1 Sym8 bior1.1 bior2.2 bior2.4 dmey

4.0 3.25 4.0 3.76 3.54 1.26

20.57 25.32 20.57 21.88 23.22 65.27

Transmission Time Analysis

Table III shows the transmission time analysis of the system for different wavelet functions employed in the encoding and decoding with same transmission scenario as used in energy analysis. Table III Comparative analysis of transmission time of 128×128 image at 230 kbps for different wavelets Wavelet Haar db1 db4 db10 coif1 coif4

Transmission time (Second) 0.41 0.43 0.46 0.56 0.46 0.59

Wavelet Sym1 Sym8 bior1.1 bior2.2 bior2.4 dmey

Transmission time (Second) 0.43 0.53 0.43 0.46 0.48 1.36

From the table it is quite clear that least amount of transmission time is achieved in case of employing Haar wavelet functions in the encoding and decoding modules of the sensor node

[449]

Farooq et al /COMMUNE – 2015

5. Conclusion In this work encoder and decoder module of a sensor node employing wavelet transform technique was implemented and investigated for different wavelets in MATLAB. Based on the experimental study of image data transfer from source node to destination node, the most suitable mother wavelet was found and it was Haar wavelet, which proved to be the best option as compared to other wavelets, which were investigated in this work. The study revealed that minimum energy consumption and least amount of transmission time for data transfer between the two nodes were achieved when Haar wavelet was employed in the encoding and decoding module. In future, this work could be extended for multimedia wireless sensor networks (MWSNs) and similar type of analysis could be made for packet losses, memory usage and execution time. References A., C.M. Decusatis and P.K. Das, Wavelets and Subbands: fundamentals and application, 1995. Graps, “An Introduction to Wavelets”, IEEE Computational Science & Engineering Magazine, Vol. 2, June 1995, pp. 50 –61. Alexandre Ciancio and Antonio Ortega, “A Distributed Wavelet Compression Algorithm for Wireless Sensor Networks using Lifting,” Proceedings of the International Conference on Acoustics, Speech and Signal Processing, Canada, May 2004. Antonini, Barlaud, Mathieu and Daubechies, “Image Coding using Wavelet Transform,” IEEE Transactions on Image Processing, April 1992, pp. 205-220. Asadollah Shahbahrami, Ramin Bahrampour, Mobin Sabbaghi, Mostafa Ayoubi, “Evaluation of Huffman and Arithmetic Algorithms for Multimedia Compression Standards,” International Journal of Computer Science, Engineering and Applications, 2011, pp. 34–47. D VijendraBabu, N R Alamelu, P Subramanian, N Ravikannan, “EBCOT using Energy Efficient Wavelet Transform,” International Conference on Computing, Communication and Networking (lCCCN), 2008. G. Anastasi, M. Conti, M. Francesco, A. Passarella, “Energy Conservation in Wireless Sensor Networks: a Survey,” Ad-Hoc Networks, Vo1.7, Issue 3, 2009, pp. 537-568. G.J. Pottie and W.J. Kaiser, “Wireless Integrated Network Sensors”, Communications of the ACM, 2000, pp. 551–558. Ian F. Akyildiz, Weilian Su, Yogesh Sankarasubramaniam and Erdal Cayirci, “A Survey on Sensor Networks,” IEEE Communications Magazine, August 2002, pp. 102-114. Karl H, Willig A, “Protocols and Architectures for Wireless Sensor Networks,” John Wiley publishing, 2005. Lee JH and Jung IB, “Adaptive Compression Based Congestion Control Technique for Wireless Sensor Networks,” Sensors (Basel), 2010, pp. 2919– 2945. Li Guang-yu, Ye Si-yuan, Zhang Zheng-xian and Gao Zong-jun, “Study and Protection of Wetland,” Marine Geology Letters, 2005, pp. 8-11. Mallat, “A Wavelet Tour of Signal Processing,” 2nd ed. Academic Press, 1999. Mohsen Nasri, AbdelHamid Helali, Halim Sghaier and Hassen Maaref, “Energy Efficient Wavelet Image Compression in Wireless Sensor Network,” International Conference on Wireless and Ubiquitous Systems, Tunisia, 8-10 October 2010. Senthil Shanmugasundaram and Robert Lourdusamy, “A Comparative Study Of Text Compression Algorithms,” International Journal of Wisdom Based Computing, Vol. 1, December 2011, pp. 68-76. Umar Farooq, Jyoti Saxena and Shabir ahmad Sofi, “Wavelet Transform based Effective Energy Utilisation Approaches of Data Transfer in Wireless Sensor Networks: A Survey,” Proceedings of ICAET, Roorke, India, May 2014, pp. 599-604. V. Mhatre and C. Rosenberg, “Design Guidelines for Wireless Sensor Networks: Communication, Clustering and Aggregation,” Ad Hoc Networks, vol. 2, No. 1, Jan. 2004, pp. 45-63. Vincent Lecuire, Cristian Duran Faundez and Nicolas Krommenacker, “Energy Efficient Transmission of Wavelet Based Images in Wireless Sensor Networks,” Eurasip journal on Image and Video Processing, 2007.

[450]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Search Interfaces of Select Online Public Access Catalogues: An Assessment Huma Shafiq*, Sheikh Mohammad Shafi, Shazia Rasool, Tariq Shafi, Zahid Ashraf Wani Department of Library & Information Science, University of Kashmir, Hazratbal, Srinagar, India

Abstract Purpose –This study is an attempt to explore and assess the search interfaces of select online public access catalogs viz WorldCat, Library of Congress Online Catalog, and NLM’s LocatorPlus. Design/methodology/approach – Due to the adoption of various information & communication, technologies by libraries in the late 80's and early 90's world over, especially in developed countries like US, the search and retrieval paradigm of library resources by users have changed considerably. keeping this as the background, the authors have undertook this study to assess and explore the new possibilities opened up by the revolutionary breakthroughs in ICT which have provided the users with new vistas for searching, retrieving and using the resources offered by the libraries to its users. To assess the changing trends in locating, searching, accessing and retrieving information, the three prominent Online Public Access Catalogs (OPACs) from Libraries in United States of America are selected for the study (WorldCat, Library of Congress Online Catalog and National Library of Medicine's (NLM’s) LocatorPlus) because of the technologically advanced nature of the users served by these famous libraries. Search Interfaces of the select OPACs are explored and assessed based on six parameters: Query Formulation, Search Limits, Help Mechanism, Alerting Services, Result Manipulation, and Additional Search Features. Findings – After assessing the search interfaces on the basis of six parameters, it is found that among the three OPACs, only Library of Congress Online Catalog has at least proved better than the other two (WorldCat and LocatorPlus) in terms of search interface features. It is apparent from the study that there is a great need to improve the search interfaces of these select OPACs as the results are not too promising due to the lack of some basic and important information retrieval features. Originality/value – The paper makes an endeavor to explore and assess the search interfaces of select Online Public Access Catalogs used worldwide.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Information Retrieval; Search Interface; Online Public Access Catalogs; WorldCat; NLM’s; LocatorPlus; Congress Online Catalog

1. Introduction An enormous amount of information is being produced and published every day. This information, which is as difficult to retrieve as is to manage, is what is called as information explosion. Whenever a piece of information is to be retrieved, a large number of results are obtained. Some of this information is pertinent to the needs of the user while some of it is a mere noise (Ryder, 2011). In order to retrieve the useful information effectively and efficiently, there must be a proper knowledge of how to retrieve the relevant information (Kules, Wilson & Shneiderman, n.d). The role of the search user interface is to aid in the searcher’s understanding and expression of their information needs, and to help users formulate their queries, select among available information sources, understand search results, and keep track of the progress of their search (Hearst, 2011). Each search interface varies in its content, presentation style and query proficiencies (He, Meng, Yu & Wu, n.d). With the intention of creating more effective search interfaces, human computer interaction experts and web designers have been developing novel interactions and features that enable users to conveniently visualize, parse, manipulate, and organize their Web search results (Wilson, Kules, Schraefel &

* Corresponding author. Tel.: +91 9622 909266. E-mail address: [email protected].

ISBN: 978-93-82288-63-3

Shafiq et al/COMMUNE – 2015

Shneiderman, 2010). An efficient information retrieval is based on the knowledge of search interfaces of the information resources and the proper use of these search interfaces. Search interfaces can be either simple or complex. For a past decade now, there has been many considerable changes in the ways and methods adopted by users to acquire the information they need. Due to the obvious outburst of information and the omnipresence of computer access, it has become much easier for the academic community to have access to this significant amount of information at their homes without personally going to the library. A vast number of online resources become available on a single click (Hiller, 2002). It is obvious that every online resource has different search interface. The present study is based on the exploration and evaluation of these search interfaces of select Online Public Access Catalogs. An Online Public Access Catalog (OPAC) is an online storehouse holding the list of all the documents available in the library or even group of libraries. OPAC helps users to search all types of documents (like Journals, Digital Materials, Books, etc.). Users primarily search OPACs to locate the exact position of documents in the library (Online Public Access Catalog, 2014). 2. Review of Literature Internet primarily started as a communication tool and now has advanced and changed significantly over the years thus, turns out to be a very vital information resource (Brinkley & Burke, 1995). Various studies, including (Berazi, 1981), (Stonier, 1991), (Aina, 2004) and (Bello and Ajala, 2004) provide proofs of the emergence of information as the most important factor in modern industrial systems (Nnadozie, 2008). New knowledge is most often produced by combining information from different sources. To produce knowledge, therefore, requires information retrieval skills. An information retrieval skill is defined as the ability to find information in such a way that non-relevant data (noise) are excluded while relevant information is found (Wien, 2000). The information that is accessible and the array of topics covered have grown to massive proportions (Brinkley & Burke, 1995; Vickery & Vickery, 1993). It has become a growing concern for users at their individual levels as well as at the organizational levels to retrieve and represent related data from enormous amount of information reservoirs. Manual retrieval systems have become outdated and unproductive to a large extent due to effective information access and sharing all over the globe with no limitation to organizational or even geographical boundaries. Thus, there is an immense necessity for effective IT-based retrieval system support (Hu, Ma & Chau, 1999). OPAC’s interface and its searching and retrieving abilities together act as a gateway to library resources and determines the user’s success in retrieving the required information. Users recognize, select and obtain the library resources with the help of bibliographic information provided by OPACs. Thus, the effectiveness in OPAC’s bibliographic display is essential for the users in order to retrieve the precise information (Mi & Weng, 2008). These OPACs started appearing in the later stage of 1990s. Since then, various libraries have implemented these and many others are considering their implementation. These online catalogues prove to be more advanced than the traditional ones due to innumerable features particularly its potential to assimilate a number of document types and sources through a single interface and in terms of providing remote access to the users (Babu & O’Brien, 2000). The way traditional catalogues used to be accessed has been totally changed over the years with the advent of a new search tool called as OPAC. It provides a number of different techniques for searching the same data, thus adding a layer of functionality. It acts as an inherently rich tool offering vast number of search features as compared to traditional catalogues (like card catalogue). It is capable of providing fast, easy and enhanced access to the users even from the remote areas hence saving their time, besides incorporating all the information of library including circulation or new arrivals (Sridhar, 2004). A study discussing library OPACs divulge that OPACs can be practically more useful by reducing the potential complexity of information to a certain manageable degree of simplicity (Wells, 2007). Many studies have been done which illustrate that OPAC’s default search options can affect users while retrieving information, thus there is a need to provide customizable search facilities in OPACs. Another study on OPACs such as Ohio State University’s Gateway and Okapi reveals that the improvements in OPACs are very promising as these systems facilitate users to their effective use without having any knowledge of resources recorded in the catalog or even without any support from experts (Thompson, Pask, Peterson & Haynes, 1994). 3. Data Analysis and Interpretation The three Online Public Access Catalogs (OPACs) are selected from United States of America (USA) and are prominently used worldwide. The OPACs are assessed on the basis of six parameters namely; (1) Query Formulation, (2) Search Limits, (3) Help Mechanism, (4) Alerting Services, (5) Result Manipulation, and (6) Additional Features. 3.1 Query Formulation Query formulation involves checking of search levels (Simple and Advanced Search) and search techniques (Keyword, Phrase, Boolean, Truncation, etc.) supported by the interfaces of the select OPAC's. All the OPACs are analyzed on the basis of these levels and techniques of query formulation. It is evident from Table 1 that under search levels both simple and advanced search facilities are available in all catalogs. While analyzing search techniques, six

[452]

Shafiq et al/COMMUNE – 2015

different search techniques are recognized. Among the six techniques, WorldCat and Library of Congress Online Catalogue support five of them (Keyword, Phrase, Truncation, Boolean and Parentheses) while LocatorPlus supports only three techniques (Keyword, Phrase and Boolean Search). ‘Proximity Operators’ technique is not supported by any of the three catalogues. Table 1: Query Formulation

Levels

Search Techniques

Simple Search Advanced Search Keyword Search Phrase Search Truncation Search Boolean Operators Parentheses Proximity Operators

WorldCat

LocatorPlus (NLM)

1 1 1 1 1 1 1 0

1 1 1 1 0 1 0 0

Library of Congress Online Catalog 1 1 1 1 1 1 1 0

*1 stands for AVAILABLE, 0 stands for NOT AVAILABLE

3.2 Search Limits While accessing the select OPAC's for retrieving the desired information, a large number of results are obtained pertaining to a particular query. In order to retrieve the most relevant information, the results need to be limited and these results can be refined by using various search limiting techniques provided by many information retrieval tools. Table 2 presents an assessment of OPACs on the basis of such search limiters. The OPACs were analyzed through ten different search limiters as listed in Table 2. A maximum of seven (Year, Material Type, Format, Location, Language, Publication Status, and Place of Publication) are supported by LocatorPlus while WorldCat and Library of Congress Online Catalogue support only five of them, (Year, Format, Language, Audience, and Content) and (Year, Material Type, Location, Language, and Place of Publication) respectively. ‘Publication Date’ search limit is offered by none of the catalogues. Table 2: Search Limits Search Limiters

WorldCat

Year Material Type Format Location Language Publication Status Place of Publication Publication Date Audience Content

LocatorPlus (NLM)

1 0 1 0 1 0 0 0 1 1

1 1 1 1 1 1 1 0 0 0

Library of Congress Online Catalogue 1 1 0 1 1 0 1 0 0 0

*1 stands for AVAILABLE, 0 stands for NOT AVAILABLE

3.3 Help Mechanism Help section in any information retrieval tool aids the user in searching the relevant information efficiently and effectively. Almost all information retrieval tools offer this facility. The three OPACs are assessed on the basis of various parameters of help mechanism. It is evident from Table 3 that six different help options exist in the OPACs under study. Library of Congress Online Catalog supports five help options (Search Assistance/Guide, F.A.Q., Feedback Facility, Contact Facility, and Instant Chat). WorldCat supports three of them (Search Assistance/Guide, Feedback Facility, and Online Tutorials) and LocatorPlus supports the least number of help options i.e. only two (Search Assistance/Guide, and Online Tutorials). Table 3: Help Mechanism WorldCat Search Assistance/Guide F.A.Q. Feedback Facility Contact Facility Instant Chat Online Tutorials

LocatorPlus (NLM)

1 0 1 0 0 1

1 0 0 0 0 1

*1 stands for AVAILABLE, 0 stands for NOT AVAILABLE

[453]

Library of Congress Online Catalog 1 1 1 1 1 0

Shafiq et al/COMMUNE – 2015

3.4 Alerting Services Alerting service keeps the user up-to-date about the latest information of his/her interest that has been added to the resource. Table 4 compares these select OPACs on the basis of various alerting services provided by them. Out of the three alerting services, Library of Congress Online Catalog supports two types (RSS Feed, and Email), and WorldCat supports only one (Email) while LocatorPlus does not support any of the three alerting services. One of the alerting services ‘Atoms Feed’ is not supported by any of the OPACs. Table 4: Alerting Service WorldCat

LocatorPlus (NLM)

0 1 0

0 0 0

RSS Feed Email Atoms Feed

Library of Congress Online Catalog 1 1 0

*1 stands for AVAILABLE, 0 stands for NOT AVAILABLE

3.5 Result Manipulation Result manipulation involves controlling the results as per the convenience of the user like reducing the results displayed per page, or sorting the results as required. The selected OPACs are evaluated on the basis of result manipulation options listed in Table 5. It is evident from Table 5 that LocatorPlus supports both result manipulation options while the other two catalogs only support one of them, WorldCat supports ‘Sorting Options’ facility and Library of Congress Online Catalog supports ‘Result Display per page’ feature. Table 5: Result Manipulation

Results Display per page

WorldCat 0

Sorting Options

LocatorPlus (NLM) 1

1

1

Library of Congress Online Catalog 1 0

*1 stands for AVAILABLE, 0 stands for NOT AVAILABLE

3.6 Additional Search Features Additional search features include all other features (e.g. Quick Search, Allied Links) offered by the OPACs that are not included in any of the five parameters of search interfaces analyzed above. Table 6 compares OPACs on the basis of these additional search features. It is depicted in Table 6 that two additional search features (Quick Search and Allied Links) have been analyzed. All the three catalogs support both these features. Table 6: Additional Search Features WorldCat

LocatorPlus (NLM)

Quick Search 1 Allied Links 1 *1 stands for AVAILABLE, 0 stands for NOT AVAILABLE

1 1

Library of Congress Online Catalog 1 1

4. Findings The major findings are enumerated as follows:  All the three OPACs provide the facility of both the levels of searching and most of the search techniques.  Most of the search limiting options can be seen in all the three Online Public Access Catalogs, thus providing the users with the facility of customizing their search according to their needs.  It is important that the users should be provided with information feedback, support and guidance in order to understand and use an information retrieval system efficiently (Park & Lim, 1999). While analyzing the help mechanism, it is not too convincing except for the Library of Congress Online Catalog.  Alerting services are also not too promising except for the Library of Congress Online Catalog again.  Result Manipulation options can be seen in all the three catalogs.  All the three OPACs fully support both the identified additional search facilities (Quick Search and Allied Links). Thus, assessing each resource on various parameters of search, it is found that Library of Congress Online Catalog provides many of the search features as compared to the rest. The other two OPACs do not show substantial results. It is clear from the above assessment that all the resources under study are in need of improving their interfaces. In order to enhance the effective, efficient and satisfactory use of application, it is important to design interface signs for the users

[454]

Shafiq et al/COMMUNE – 2015

(Islam, 2013). It can thus be concluded that online resources have to be designed in such a way that the users can make the most from their use. 5. Conclusion Gone are the olden days when a user had to pass through a cumbersome task of searching the various catalogue cabinets housed at a separate place in the library for locating a desired book. OPAC's have now revolutionized the way information is being located and searched in the libraries. Developing the more user friendly OPACs is the first step in ensuring that the users of a particular library are satisfied upto the optimum level and it can go a long way in saving the time and reducing the frustration of the users in locating their desired documents in the library. It is the need of the hour and foremost responsibility of the librarians to develop the OPACs as per the requirements of the users they serve, as the complex nature of the information sources acquired by the libraries of the twenty first century are changing considerably. References Babu, B.R., O’Brien, A. 2000. Web OPAC interfaces: An overview. The Electronic Library, 18(5), 316-330. DOI: 10.1108/02640470010354572 Brinkley, M., Burke, M. 1995. Information retrieval from the internet: An evaluation of the tools. Internet Research: Electronic Networking Applications and Policy, 5(3), 3-10. DOI: 10.1108/10662249510104595 He, H., Meng, W., Yu, C., Wu, Z. n.d. Construction Interface schemas for Search Interfaces of Web Databases. Retrieved from: http://cs.binghamton.edu/~meng/pub.d/He_p136.pdf Hearst, M.A. 2011. User Interfaces for Search. Modern Information Retrieval: The Concepts and Technology behind Search Engines (2nd ed.). Boston, USA: Addison Wesley Professional. Retrieved from: http://people.ischool.berkeley.edu/~hearst/papers/mir2e_chapter2_hearst_uis_references.pdf Hiller, S. 2002. The impact of information technology and online library resources on research, teaching and library use at the University Of Washington. Performance Measurement and Metrics, 3(3), 134-139. DOI: 10.1108/14678040210454923 Hu, P.J.H., Ma, P.C., Chau, P.Y.K. 1999. Evaluation of user interface designs for information retrieval systems: A computer-based experiment. Decision Support Systems, 27(1-2), 125-143. DOI: 10.1016/S0167-9236(99)00040-8 Islam, M.N. 2013. A systematic literature review of semiotics perception in user interfaces. Journal of Systems and Information Technology, 15(1), 45-77. DOI: 10.1108/13287261311322585 Kules, B., Wilson, M.L., Shneiderman, B. n.d. From Keyword Search to Exploration: How Result Visualization Aids Discovery on the Web. Retrieved from: http://hcil2.cs.umd.edu/trs/2008-06/2008-06.pdf Mi, J., Weng, C. 2008. Revitalizing the library OPAC: Interface, Searching, and Display Challenges. Information Technology and Libraries, 27(1), 522. DOI: 10.6017/ital.v27i1.3259 Nnadozie, C.O. 2008. Use and Non-Use of Information Resources by Low-Ranking Industrial Employees in Nigeria. Trends in Information Management, 4(2), 140-157. Retrieved from: http://ojs.uok.edu.in/ojs/index.php/crdr/article/view/130/118 Online Public Access Catalog, 2014. Retrieved from: http://en.wikipedia.org/wiki/Online_public_access_catalog, Jun. 19, 2014 [Oct. 5, 2014] Park, K.S., Lim, C.H. 1999. A structured methodology for comparative evaluation of user interface designs using usability criteria and measures. International Journal of Industrial Ergonomics, 23(5-6), 379-389.Retrieved from: http://ac.els-cdn.com/S0169814197000590/1-s2.0-S0169814197000590-main.pdf?_tid=622575ba-b482-11e2-8366 00000aab0f01&acdnat=1367648466_02ab7aebb51b8350b86caa7a82e9cb40 Ryder, B. 2011, January 30. Too much information. The Economist. Retrieved from: http://www.economist.com/node/18895468#footnote2 Sridhar, M.S. 2004. OPAC vs card catalogue: A comparative study of user behaviour. The Electronic Library, 22(2), 175-183. DOI: 10.1108/02640470410533443 Thompson, D.M., Pask, J., Peterson, B., Haynes, E. 1994. Online public access catalogs and user instruction. Reference & User Services Quarterly, 34(2), 191-202. Retrieved from: http://www.jstor.org/stable/pdfplus/20862644.pdf Vickery, B., Vickery, A. 1993. Online search interface design. Journal of Documentation, 49(2). Retrieved from: http://www.emeraldinsight.com/search.htm?st1=Online+search+interface+design&ct=all&ec=1&bf=1 Wells, D. 2007. What is a library OPAC?. The Electronic Library, 25(4), 386-394. DOI: 10.1108/02640470710779790 Wien, C. 2000. Teaching online information retrieval to students of journalism. Aslib Proceedings, 52(1), 39-47. DOI: 10.1108/EUM0000000006999 Wilson, M.L., Kules, B., Schraefel, M.C., Shneiderman, B. 2010. From Keyword Search to Exploration: Designing Future Search Interfaces for the Web. Foundations and Trends in Web Science, 2(1), 1-97. DOI: 10.1561/1800000003

[455]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Big Data: A Growing Tide not Hype Samiah Jan Nastia*, M. Asgarb , Muheet Ahmed Buttc, Majid Zaman Babad a Department of Computer Sciences, BGSB University, Rajouri, India Department of Mathematical Sciences and Engineering, BGSB University, Rajouri, India c Department of Computer Sciences, University of Kashmir, Srinagar, India d Directorate of IT and SS, University of Kashmir, Srinagar, India

b

Abstract Data revolution is just at its infancy; everyone is talking about Big Data. Big Data is an explosion of data and as such traditional systems are not scalable enough to handle this enormous data. The explosion of the Big Data is a very recent phenomena it is quite recently, companies have started to realize that they should capture all this data that is being producing and not only capture they should try to analyse it and try to get some value of it. This paper explores the Sources of Big Data, Architecture of Big Data, challenges and issues produced by it and Hadoop, a Big Data tool

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Big Data: Architecture: Hadoop

1. Introduction Data is a new class of economic asset, like currency and gold (World Economic Forum 2012). Data is growing at enormous rate, so it is very difficult to manage and handle this huge and gigantic volume of data. It is very difficult to handle this enormous data because it is growing so rapidly in comparison to the computing resources. The term Big Data is very confusing as it gives us a feeling that after a certain size the data is big and below a certain size the data is small (Dong, X.L and .Srivastava, D). The Big Data could start from any point. There is no definitive definition for Big Data. However it is mostly defined this way that “Big Data is a data that becomes difficult to be processed because of its size using traditional system”. Traditional systems including relational databases are not capable of handling the Big Data and challenges spring up at multiple levels including capturing, storing, analysing, searching, sharing, transforming the data and even visualizing the data. The Big Data becomes a challenge for traditional systems not merely because of its size that could be a challenging point but challenge may also arise because of its speed at which the Big Data is coming in and also because it is unstructured and it could contain data items of various formats. So Big Data is usually measured by three attributes, velocity, volume and variety. The velocity refers to the speed at which the data is coming in e.g. the Scientific Experiments that they do at atomic reactors where they do the collision of sub-atomic particles, 40TB of data could come in within one sec, so that is a very high speed. Volume is of course a problem, the data keeps on accumulating and the file becomes too large to be handled by traditional system. The Facebook is generating 25TB of data daily so just imagine the size of the files that are there since the beginning of time. In traditional systems data is structured and is stored well in planned tables, each table has specific columns and each column could accept values of specific data types. However in case of Big Data, the third V creates problem sometimes i.e. variety. When the Big Data comes in it may include items of variety of formats. It could have audio files, video files, and unstructured data like text messages so that becomes challenging sometimes for a traditional system to handle. The explosion of the big data is a very recent phenomenon and it is quite recently, companies have started to realize that they should capture all this data that is being produced and not only capture they should try to analyse it and try to get some value out of it. These days the decision making is solely performed on structured data which is mostly stored in applications like ERP’s and other related applications that are running in an Enterprise. So, the most of this unstructured data gets wasted, it is not captured and *

Corresponding author. Tel.: +91 8491 027772. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Samiah et al/COMMUNE – 2015

even if it is captured it is not analysed and even if an attempt is made to analyse this data, the real value is not expected because of the limitations. This is the area (90% area) that represents the focus of the companies in coming years and these companies will try to analyse this unstructured data and extract meaningful information out of it. 2. Related Work Kaisler and his Co-workers analyse the issues and challenges in Big Data and begin a collaborative research work program into methodologies for Big Data analysis and design. They conclude that the traditional databases do not solve the various aspects of the Big Data problem and thus advocates for machine learning algorithms which should be more robust and easier for unsophisticated users to apply. Bakshi. K discusses architecture of Big Data and ends up with the conclusion that despite the various architectures and design decisions the analytics system aim to scale out elasticity and high availability. The concepts of Big Data and the available market solutions used to manage and explore the unstructured data are discussed. The results and observations thereof concluded that the analytics is a pivotal part for adding value for the social business. Demchenko , Zhiming and Wibison introduces the universal model known as the Scientific Data Infrastructure (SDI).Authors show that with the help of Cloud based Infrastructure such a model can be implemented easily. Courtney, M investigates the difference in Big data applications and how they are different from the traditional methods of analytics existing from a long time. Smith and his co-workers analyze social media sites such as Flickr, Locr, Facebook and Google+. Based on this analysis they have discussed the privacy implications and also geo-tagged social media; an emerging trend in social media sites. They presented a concept with which users can stay informed about which parts of the social media deluge relevant to them. 3. Sources of Big Data A growing number of users, applications, systems, sensors, mobile devices, social media etc, are producing large and large files. These files are not only large, they are being produced at a very high speed and sometimes these file contain variety of data items which are not even structured such as video files, audio files, images, photos, logs, click trails, text messages, emails documents, books, transactions, public records etc. So all these attributes are creating challenges for traditional systems and hence the term Big Data. Here are some examples of data generation points: (Katal, A. et al) 3.1

Data from Enterprises

Now –a- days the profitability of business department is mainly influenced by the IT and digital data. It has been estimated that the data used and produced by the different business companies may double every 14 months. Amazon terminal operations are processed in millions and third party sellers’ queries are more than 500K per day .Also1 million customer trades are processed by Wal-Mart per hour. Big Data is also coming from activities like Trading. New York Stock Exchange (NYSE) produces 1TB per trading day; in 2020 the total data estimated is 35 ZB. 3.2

Internet of Things

IoT is a vital source of Big Data. The cities which have been designed on the basis of IoT are referred to as Smart Cities and in these cities the data comes from various fields such as medical care, Agriculture, Traffic, Industry etc. The data generated by the IoT has mainly the following features: i. Large Scale data, ii. Heterogeneity of Data , iii. Time and Space Co-relation. 3.3

Biomedical Data

The human genome project (HGP) and the sequencing technology lead to a generation of huge amount of data. A data of about 100-600 GB is generated by one sequencing of human gene. By 2015 the traceable biological samples will reach to 30 million. The development of gene sequencing and bio-medical technologies will contribute to the continuous growth of Big Data of Bio-medicine beyond all doubts. 3.4

Other Sources

Airbus generates 10 TB every 30 minutes. About 640 TB is generated in one flight. This is a lot of data usually one warehouse would be of this size .Smart meters read the usage every 15 minutes and record 350 billion transactions in a year. By the end of 2014, there were 200 million smart meters. Camera phones are the world wide and most of them are

[457]

Samiah et al/COMMUNE – 2015

location aware meaning when we take a photo the location information could go in the photo as well as because of the GPS capabilities and 22% of these phones are smart phones. By the end of 2013, it was estimated that the number of smart phones exceed the number of PC’s. The cell phones and smart phones are major players in creating large volume of data. According to CISCO it was estimated that by 2014, internet traffic was reached 4.8 Zeta Bytes. Sloan digital sky survey (SDSS) and High energy physics also are the major sources of Big Data 4. Architecture of Big Data Big Data resembles to the Cloud Architectures and is having four layers as shown below in figure 1:

Fig 1: Cloud Architectures

Infra-structure as a service: - It consists of network as a base, storage, servers and inexpensive commodities of Big Data stack which can be bare metal or virtual (cloud). Distributed file system are part of this layer. Platform as a service: - Distributed caches and NoSQL stores, form the platform layer of the Big Data. The logical model is provided by this layer for the unstructured and raw data stored in the files. Data as a service layer: - In this layer the integration of all the tools available is done with the PaaS Layer using integration adapters, search engines, Batch programs etc. Big Data function as a service: Packaged applications can be built by the industries like e-commerce, health, retail and banking to serve a specific business need and the DaaS layer is leveraged for the cross cutting data functions. 5. Challenges and Issues in Big Data The challenges and issues in big data include (Madden, S and Wand W): 5.1

Information Sharing and data access

In order to make timely and accurate decisions it is mandatory that the data should be timely available and besides its timely availability it should be also complete and accurate. This necessity of making data open and available to government agencies in standardized manner leads to decision making , Business intelligence and productivity improvements but this process of making data open makes the management and governance process bit complex and challenging. In order to get an edge in business the sharing of data by the companies about their operations and clients, threatens the intrinsic of secrecy and competitiveness. So it is very much akward to expect the sharing of data between the companies. 5.2

Storage Issues

The large amount of data generated by almost everything such as social media sites, sensor devices etc. needs to be stored and the storage available is not enough for such amount of data. Uploading such large amount of data in Cloud seem an option but it does not solve the problem in actual. The data which is in Tera Bytes make take a large amount of time for uploading in Cloud and it is hard to upload this rapidly changing data in real time. Besides the above the Cloud distributed nature is a hurdle of analysis of Big Data. This outsourcing this Big Data to cloud leads to Capacity and Performance Issues. 5.3

Processing Issues

It takes large amount of time for processing such large amount of data. Whole of data set is required to be scanned to find suitable elements somewhat which is not possible and building up of indexes right at the start while storing and collecting the data is a good option as it reduces time of processing. It considerably reduces analytical challenges.

[458]

Samiah et al/COMMUNE – 2015

5.5

Analytical Challenge

This huge and massive amount of data which can be structured, semi-structured and structured requires skills for its analysis. Moreover the analysis to be done on the data depends highly on the results to be obtained i.e. decision making, and this can be done either by incorporating massive data volumes in analysis or determine upfront which Big Data is relevant. The main challenges can be summarized as:      5.6

What, When data volume gets varied and large and how to deal with as it is not known? How data can be used to best advantage? Is it necessary to store all data? Is it necessary to analyse all the data? How to search or find out data points which are really important?

Skill Requirement

Big Data is a new and emerging technology and needs to attract the organizations and people with diverse new skill sets, which should not only include technical but also should extent to other areas such as research analytical and creative ones. 6. Hadoop: A Big Data Tool Hadoop – an open source framework was developed by Google and later on adopted by Yahoo and handed over to Apache. Hadoop supports the processing of large data sets in a distributed environment. It breaks the data into smaller pieces and thus breaks the computation into smaller pieces as well as each smaller piece of computation is sent to the smaller piece of data so that instead of performing on Big computation, numerous smaller computations are performed obviously much faster and finally the result is aggregated and the aggregated result is sent back to the application. There are two core components of Hadoop viz. Hadoop File Distributed File System, HDFS (Storage) and Map Reduce (Processing) as shown below in Figure 2

Fig. 2: Hadoop File Distributed File System

7. Conclusion Everything we do generates data. We swim in such a sea of data whose level is increasing. Millions and Billions of people and sensors and trillions of transactions are rapidly working to generate unimaginable amounts of data without any doubt. Technology evolution and placement guarantee that in few years more data will be available in year than has been created since the dawn of the man. References Bakshi, k., 2012. Considerations for Big Data: Architecture and Approach, IEEE , Aerospace Conference, p.1 Courtney, M., 2012. The Larging-up of Big Data, IEEE, Engineering &Technology,p.72 Demchenko, Y, Zhiming Zhao ; Grosso, P., Wibisono, A., de Laat, C. , 2012. Addressing Big Data Challenges for Scientific Data Infrastructure, IEEE , 4th International Conference on Cloud Computing Technology and Science,p.614 Dong, X.L. , Srivastava, D. , 2013. Big Data Integration, IEEE 29th International Conference on Data Engineering (ICDE), p. 1245 Katal, A., Wazid, M., Goudar, R.H. , 2013. Big data: Issues, challenges, tools and Good practices , Sixth International Conference on Contemporary Computing (IC3) , p. 6 Kaisler, S., Armour, F., Espinosa, J.A. ,Money, W. , 2013. Big Data : Issues and Challenges Moving Forward, IEEE, 46th Hawaii International Conference on System Sciences, p.995 Madden, S ., 2012 . From Databases to Big Data, IEEE, Internet Computing,p.4 Singh, S. ; Singh, N., 2012. Big Data Analytics, IEEE, International Conference on Communication, Information & Computing Technology (ICCICT),p.19 Smith, M., Szongott, C., Henne, B., von Voigt, G. , 2012. Big Data Privacy Issues in Public Social Media, IEEE, 6th International Conference on Digital Ecosystems Technologies (DEST), p.1 Wang, W , 2014 .Big Data , Big Challenges ,, IEEE International Conference on Semantic Computing (ICSC) ,p. 6

[459]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

English-Kashmiri Machine Translation: Issues and Challenges Mir Aadila*, Mohammad Asgerb, Vishal Goyalc a Department of Computer Sciences, BGSBU, Rajouri, India. School of Engineering & Mathematics, BGSBU, Rajouri, India c Department of Computer Sciences, Panjabi University, Patiala, India b

Abstract Machine Translation is now considered as a challenging task for research by the academicians. It is an interesting and promising study of research, even though a flawless and correct translation by an intelligent computer is yet a dream to be realized due to the complexity and challenges that slowly came to notice. Most of these problems are independent of the methodology or tools used to achieve overall translation but still vary with each language pair. Every language pair puts forth a different level of challenges and issues which latter on becomes the reason of undesirable translation quality and fluency. This work tries to bring forth few of the main challenges that are faced at the very initial stages of the process of machine translation for English-Kashmiri machine translation.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Machine Translation; English-Kashmiri Translation; Challenges with Machine Translation; Divergence in English-Kashmiri

1. Introduction Machine translation is not merely an automatic linguistic word by word translation, rather a translation of one natural language to another and preserving the meaning just like a human translator. And just like a human translator, machine translators also face multiple divergences in any language pair. The languages differ in their lexicon, syntax, semantics, pragmatics, culture and background so these need to be taken in account for a reliable translation. Kashmiri culture, its beauty, its Sufi saints and its literature is diverse and unique and so is its language. However, the language (like other around 3000 languages) is facing a treat of extinction. That is why globalization of its language is the need of the hour. Machine translation is a promising solution. But Kashmiri language has a really scarce corpus available online or in digital form. And the development of a huge database of parallel corpora is one of the biggest challenges. Also like any language pair, English-Kashmiri translation also exhibits a general and an idiosyncratic difference in realization of their syntax and word order and has a lexical and morphological divergence. These divergences put forth some issues and challenges, most of which creep in usually the initial stages of machine translation and continuously damage the overall efficiency of the translation. The main challenges and issues that arise are a result of the ambiguity in source language, divergence across the languages and finally the variations in the target language. A study of these root causes is necessary as these are to be kept in view while devising algorithms for machine translation. 2. Source Language Ambiguity Nagamani and Ananth proposed an image compression technique for high resolution, grayscale Satellite urban images. The proposed technique used discrete wavelet transform together with EZW (Embedded Zero tree wavelet) and SPIHT (Set Partitioning in Hierarchical Trees) coding techniques in order to achieve high compression ratio and better image quality. The compression ratio and peak signal to noise ratio determined using EZW and SPIHT codings have been compared to each other for same set of images. The results obtained showed possibility to achieve higher * Corresponding author. Tel.: +91 9086 750369. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Aadil et al/ COMMUNE-2015

compression ratio and PSNR (approximately CR of 8 and PSNR of 29.20) for SPIHT coding compared to EZW coding (approximately CR of 1.07 and PSNR of 13.07) for applications related to satellite urban imagery (Nagamani, Ananth, 2011). Ambiguity in source language are not itself a big issue as these are usually solved in context, but what damages the efficiency of translation is that these ambiguities multiply. A 2-gram sentence(two words) with each word having only two different meanings can be interpreted in four different ways and a 3-gram with same degree of ambiguity shall have eight different interpretations all correct as per grammar. 2.1

Lexical Ambiguity.

When there are multiple meanings of same word or phrase in the source language. e.g. The word “bank” can be used for river edge as well as for a financial institution. 2.2

Syntactic Ambiguity.

When there are multiple interpretations for a sentence because of unclear modifying expression in its structure- its syntax. e.g. “We need apple or banana and sugar” may mean either apple or both banana and sugar or it may mean apple and sugar or banana and sugar. Similarly, for the sentence “It is a little pretty girl's doll” may mean the doll is little and girl is pretty or the girl is little and pretty or the doll is little and pretty. 3. Cross Lingual Divergences Divergence is the most common observation between any two languages to be translated. When two sentences of the source language that are structurally similar show divergence in their structure in the target language, such languages show cross-lingual divergence. [Dorr, 1993]. For our study we used some of the divergence types based on the Dorr Classification [1993]. Dorr categorizes translation divergence broadly in  Syntactic Divergences like Constituent order divergence, Adjunction divergence, Preposition Stranding divergence, Movement divergence, Null Subject divergence, Dative Divergence and Pleonastic Divergence.  Lexical-Semantic Divergences like Thematic Divergence, Promotional Divergence, Demotional Divergence, Conflactional/Inflational Divergence, Categorial Divergence and Lexical Divergence. Some of these divergences are universal and some exist only for particular language pairs. Some particular of these divergences in English-Kashmiri language pair are as described below. 3.1. Categorial Divergence: It is the most common type of divergence found in any language pair. Categorial divergence results when parts of speech in the source language are translated into the target language using different parts of speech, e.g. if Noun in source language is translated into adjective in the translated language or vice-versa, or if verb in source language is translated into noun in the target language or vice-versa. So, any change in the lexical category of a word in the source language during translation process leads to categorial divergence, e.g. Noun Adjective or Adjective Noun, Verb Noun or Noun Verb Some of the examples showing categorial divergences in English-Kashmiri Translation are. 

Verb to Noun: English: My mom loves me. Kashmiri: Transliteration: Maineh Majeh Che Main Maiye. Or Kashmiri: Transliteration: Maineh Majeh Che Maiye Main. In English sentence the word “loves” is a verb while in Kashmiri language it is realized as noun.



Noun to Verb: English: His residence is near the river. Kashmir: Transliteration: Su chu daryavas nazdeek roozan.

[461]

Aadil et al/ COMMUNE-2015

In English the word “residence” is a noun while in Kashmiri it is realized by verb. 

Noun to Adjective: English: She is a beauty. Kashmir: Transliteration: Sou che Khoobsurat. The word “beauty” in English is a noun while in Kashmiri it is adjective.



Adjective to Noun: English: Her behavior is harsh. Kashmir: Transliteration: Tinhindis wartravas manz chu troushar.

3.2. Conflational and Inflational Divergence: Conflational Divergence occurs when two or more words in the source language are translated in a single word by combining their meaning in the target language. Such divergence is also known as Lexical Gap. e.g. English: He slipped away. Kashmiri: Transliteration: (su) Tsul English: You may leave now Kashmir: Transliteration: Neeriv Inflational Divergence is just the opposite of the Conflational Divergence. It is when a single word in source language is translated in multiple words in the target language. English: Please leave. Kashmiri: Transliteration: Meharbani kareth neiriv toye. English: It suffices Kashmiri: Transliteration: ye chu hajtas mutabik 3.3. Structural Divergence: Structural divergence is the difference in the realization of the incorporated arguments of the source language and in the target language. This is the difference in Phrasal Categories, e.g. when an ad-position-phrase (PP) category in one language is realized by a noun-phrase (NP). English: I have to go to Punjab. Kashmiri: Transliteration: Meh chu gasun Punjab. English: I have to write a letter. Kashmiri: Transliteration: Meh che chithe lyekan 3.4. Head Swapping Divergence: Head Swapping arises when the role of the main verb in source language is diminished and the role of the modifier verb is promoted in the target language. The first phenomenon is known as demotional divergence and the second is known as promotional divergence respectively. Since in almost all cases demotional and promotional divergences work together, these are together entitled as Head Swapping Divergence. English: The Music is on. Kashmiri:

[462]

Aadil et al/ COMMUNE-2015

Transliteration: saaz chu bahjaan ya wazaan English: The Jhelum is flowing. Kashmiri: Transliteration: Wyeth che pakaan 3.5. Thematic Divergence: Thematic divergence arises due to switching of role of arguments in the Source Language to Target Language, e.g. the difference in the argument structure of verb in the two languages. English: Where from are you? Kashmiri: Transliteration: Toye kateh chiv roozan. (OR) Kashmiri: Transliteration: Toye kateh peth chiv. English: Why are you late. Kashmiri: Transliteration: Toye kyazi gov tsheer. 3.6. Lexical Divergence: Lexical divergence is not very uncommon. This is relatively simple type of divergence where exact match for translation of a word or phrase in the Source language is not available in the Target Language. The lack of exact translation map for a certain construction between two languages gives rise to such divergence. English: Good Luck Kashmiri: Transliteration: Khudaiye sund fazal aesney English: Please Sit Kashmiri: Transliteration: Meharbani Kareth thayev tashreef. 3.7. Honorific Divergence: As other South Asian languages, Kashmiri Language also exhibits some of honorific features. Honorific divergence occurs when we use plural inflectional elements (verb and the genitive noun) for some nouns known as Honorific nouns just for the purpose of showing respect or honor. Such features are not present in English language and other European languages. English: He is my friend. Kashmiri: Transliteration: Suh chu meh doost English: He is my teacher. Kashmiri: Transliteration: Tem cheh mein ustaad English: He has come. Kashmiri: Transliteration: Suh aaw (or) Timav aun tashreef.

4. Target Language Variation For a single sentence, each word can be translated into multiple options resulting in a large number of translated sentences in target language. This may result due to ambiguous structure, but the main cause is word sense disambiguation. Although for corpora rich languages like English it can be solved by anaphora resolution, coherence, and inference and mostly by improving relevance of search engines. However for a corpora deficient language like

[463]

Aadil et al/ COMMUNE-2015

Kashmiri, it is a bigger problem. Statistical Machine translation approach uses the probability and frequency of each sentence out of the multiple candidate sentences to decide the most appropriate translation. This may not always work as a sentence is usually a reflection of thought of mind and the most frequent or probable translated sentence may not be always correct. However, Kashmiri language is still not abundantly available on Internet, so resolving word sense disambiguation for Kashmiri language is typically tough. 5. Conclusion There are multiple issues and challenges in English-Kashmiri machine translation that need to addressed to ensure fluent, efficient, and desirable output. The scarcity of parallel corpora, the divergences and variations at different levels make English-Kashmiri translation difficult. These divergences and variations need to be studied properly to devise proper algorithms and tuning of the translation mechanisms for a better output. There are various techniques and methods that can resolve the issues of these divergences. However, not each one of these challenges can be solved yet. References Aasim Ali and Malik, M. K., 2010. Development of parallel corpus and english to urdu statistical machine translation. Int. J. of Engineering & Technology IJET-IJENS, 10:31–33. Jawaid, B. and Zeman, D., 2011. Word-order issues in english-to-urdu statistical machine translation. Number 95, pages 87–106, Praha, Czechia. Dave, S. and Parikh, J. and Bhattacharya, P., 2002. “Translation Technical Report", LAMP 88. Dorr, B., 1994. “Classification of Machine Translation Divergences and a Proposed Solution Computational Linguistics”. 20 (4) 597–633. Dorr. Bonnie, J., 1994. “Machine Translation Divergences: A Formal Description and Proposed Solution”, Computational Linguistics, 20:4, pp. 597-633. Dorr, Bonnie, J. and Nizar Habash,, 2002. “Interlingua Approximation: A Generation-Heavy Approach”, In Proceedings of Workshop on Interlingua Reliability, Fifth Conference of the Association for Machine Translation in the Americas, AMTA-2002,Tiburon, CA, pp. 1—6. Dorr, Bonnie J., Clare R. Voss, Eric Peterson, and Michael Kiker,. 1994. “Concept Based Lexical Selection”, Proceedings of the AAAI-94 fall symposium on Knowledge Representation for Natural Language Processing in Implemented Systems, New Orleans, LA, pp. 21—30. Dorr, Bonnie J., Lisa Pearl, Rebecca Hwa, and Nizar Habash,. 2002. “DUSTer: A Method for Unraveling Cross-Language Divergences for Statistical Word-Level Alignment," Proceedings of the Fifth Conference of the Association for Machine Translation in the Americas, AMTA-2002, Tiburon, CA, pp. 31—43. Dorr, Bonnie, J., and Nizar Habash, 2002. “Handling Translation Divergences: Combining Statistical and Symbolic Techniques in Generation-Heavy Machine Translation”, In Proceedings of the Fifth Conference of the Association for Machine Translation in the Americas, AMTA-2002, Tiburon, CA, pp. 84—93. Goyal, P., and Sinha. R.M.K., 2008. “A Study towards English to Sanskrit Machine Translation system”. SISSCL. Haspelmath, Martin. 2002. “Understanding Morphology”, Oxford University Press. Jawaid, B., Zeman, D., Bojar, O., 2010. “Statistical Machine Translation between Languages with Significant Word Order Difference”. PBML. Kameyama, Megumi and Ochitani, Stanley Peters. 1991. “Resolving Translation Mismatches With Information Flow” Annual Meeting of the Assocation of Computational Linguistics. Koehn, P., 2010. “Statistical Machine Translation”: Cambridge University Press. Levin, B., 1997. “English Verb Classes and Alterations: A Preliminary Investigation”, the MIT Press. Lewis, Paul, M., Simons, G.F., Fennig. C.D., 2013. “Ethnologue: Language of the World”. Seventeenth edition. Dallas, Texas: SILI. Sinha, RMK and Thakur, A., 2005. “Translation Divergence in English-Hindi MT EAMT”, 10th Annual Conference, Budapest, Hungary

[464]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16-18March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

A Comparative Analysis of Full Adder Cells in Nano-Scale for Cascaded Applications Afshan Amin Khan*, Shivendra Pandey, Jyotirmoy Pathak Lovely Professional University,Punjab,Jalandhar,India

Abstract This paper focuses on the different designs of an Adder cell with an aim of finding an Adder cell from the literature, which can be used for the future VLSI application. To ensure this all the designs have been implemented in Cadence Virtuoso 90nm Technology with least possible size of transistors available. A critical parameter in identifying circuits suitable for VLSI applications is that the implementation area and power dissipation must be as small as possible, thus in this paper we are trying to compare different possible design of an adder cell ranging from the most stable 28 transistor(28T) adder cell to a low area 8 transistor(8T) adder cell. All the designs have been compared for their threshold voltage loss, power dissipation, delay, and PDP, to choose the more reliable adder cell for cascaded VLSI applications.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Low Arear Full Adder; Nano Scale Adder; Lower Trasistor Count; Threshold Loss; Cascaded Logic

1. Introduction In recent years, a good number of researchers have focused their work for the optimization of system architectures such as that of a Filter, multipliers, MAC units etc. This methodology of optimization includes a huge amount of theoretical calculations and improvisations in the logic used to implement the architecture. However, a simpler method to optimize any logic circuit can be to identify a base cell that can be repeated to form the complete architectural design of a system. For systems like multipliers and others the base cell is clearly a Full Adder Cell. Thus optimization of a Full Adder Cell will lead to the optimization of the system as a whole while all the efforts are focused on optimizing a single unit out of the whole architecture. Realizing the significance of a Full Adder Cell we have gone through the literature and identified a wide range of versatility in the design and also in the number of transistors used to implement each design. Some of the promising designs of Full Adder Cells have been implemented and compared in this paper. However a disadvantage of reducing the number of transistors being used to implement the Adder cell is that the driving capability is also expected to vary with the number of transistors used to implement the cell. Thus the cells given in this paper have been compared on the basis of amount of the degradation attained in the output logic level for respective input combinations, power consumption and delay. As we tend to reduce the number of the transistors used to implement an adder cell starting from the fully symmetrical design of a 28T Adder cell R. Zimmermann et al., 1997 and going down to least possible 8T Adder cell Fayed et al., 2001 there is a variation in the amount of the degradation levels attained in either Sum or Carry output or both, which ultimately restricts the driving capability of respective Full Adder Cell. Each adder cell has been somewhat modified by varying the W/L ratio of the transistors used in the design of that particular adder cell so as to attain some acceptable logic levels of Sum and Carry outputs, which help in proper driving of the next stages of the circuit.

2. Implementation and Design Analysis A detailed study of the literature provides different possible designs of a Full Adder Cell that are optimized in one or the other respect. However, an optimization is reliable only if it fits in some real time application such as multi-bit addition, Multiplication and so on, in order to find an application suitable adder cell following designs where implemented and analyzed for the features given below.

* Corresponding author. Tel.: +91 9796 570562. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Khan et al/COMMUNE – 2015

2.1.

28T Adder Cell.

The 28T CMOS adder cell uses 28 transistors for its implementation, an efficient design of this 28T full adder R. Zimmermann et al., 1997 provides a symmetric design of 28T full adder as in Figure 1. The basic advantage of this adder is that it has no threshold loss problem and more over power dissipation is of acceptable values. As this adder is being designed using CMOS design, it stands as one of the adder that has a versatile application and will be as a stand out performer in most of the cases, whether it be the large systems or an adder cell itself, 28T adder is a promising design, hence becomes a major challenge to the researcher to optimized such a stable design. The two major concerns for the implementation of this cell is the amount of area consumed and the amount of delay of this adder cell thus has becomes an area of research for the designers. As a result of this research designers have developed different circuit level implementations which take care of reducing the implementation area and delay with some or little tradeoff of the circuit performance. However some designs not only provide an area reduction but also show a significant improvement in performance characteristics.

Fig 1. 28T CMOS Full Adder cell

2.2.

24T Adder Cell.

The 24T adder cell S. Goel et al., 2006 given in Figure 2 is also one among the promising adder cell in terms of less threshold loss problem as well as reduces the number of transistors by four, however the amount of the power dissipation in this circuit increases as compared to the most stable 28T adder cell and is nearly double than the power of the 28T adder cell, whereas the amount of area saved is very less. Hence it is a requirement for the designer to optimize this design for power consideration. Hence can’t be regarded as efficient trade-off of power and area.

Fig 2. 24T Full Adder cell

2.3.

20T Adder Cell.

This adder cell uses 20 transistors for implementing both sum and carry functions. A possible design of Adder cell using 20T N. Weste and K. Eshraghian et al., 1993 is given in Fig 3. This adder cell was also observed to have very little or no threshold loss problem, more over the number of transistors is reduced by 8. Thus area is reduced to a great deal with significant improvement in power consumption. Hence it assures to be better performer in terms of both power and area.

[466]

Khan et al/COMMUNE – 2015

Fig. 3. 20T Full Adder cell

2.4.

16T Full Adder Cell.

The 16T adder cell N. Zhuang and H. Hu., 1992 given in Fig. 4 proves to be a better design than the above designs in terms of power dissipation as well as delay is almost similar to that of 28 and 24T adder cells, whereas there is a slight degradation in output logic levels. However the driving capability is still expected to be fairly good. The design is a mixture of the pass transistor logic, GDI and pass transistor based MUX cells as in Figure 4.

Fig. 4. 16T Full Adder cell

2.5.

14T Adder Cell.

Two different designs of 14T adder cells have been taken from Chip-Hong Chang et al., 2003 and T.Sharma et al., 2010. Among these 14T adder of Chip-Hong Chang et al., 2003 proves to be a better adder and gives astonishing results in terms of voltage threshold loss problem, however the amount of power consumed increases. The one in T.Sharma et al., 2010 gives some serious threshold loss problems and even supplies some very weak logic values for the w/l ratios used to implement this design, hence is a further topic of research. In this work we have considered the 14T adder cell in Chip-Hong Chang et al., 2003 shown in Figure 5 only for the rest of analysis and comparison with other designs.

[467]

Khan et al/COMMUNE – 2015

Fig. 5. 14T Full Adder cell

2.6.

12T Full Adder Cell

The 12T adder cell Yingtao Jiang et al., 2004 given in Fig. 7 proves to be a promising design in terms of least amount of power dissipation among all the circuits implemented in this work but on the other hand the output logic voltage levels for respective logic are not satisfactory and needs some level restoration circuitry or improved sizing parameters to be used for obtaining proper logic voltage swings for both Sum and Carry outputs of a Full Adder cell.

Fig6. 12T Full Adder cell

2.7.

10T Adder Cell.

The 10T adder cell Fayed et al., 2001 given in Figure 8 proves to be very useful adder cell in terms of reducing the overall power of the circuit but as expected the more we go away from the conventional design possible chances of threshold loss problem are more as a result the output obtained is degraded in nature. Thus this adder can’t be used for a circuit where the cascading of the adder cell structures is very huge. However comparing the adder for the amount of area saved and amount of power dissipation obtained , the amount of degradation in the output is somewhat acceptable and may be able to drive some limited number of stages in a system were cascading is a choice of design.

Fig. 7. 10T Full Adder cell

[468]

Khan et al/COMMUNE – 2015

2.8.

8T Adder Cell.

The design of 8T adder cell Shiwani Singh et al., 2012 is observed one of the reliable designs among all other 8T adder cell analyzed so far in Afshan Amin et al., 2014. This cell gives the advantage of reducing the number of transistors to a great deal however its huge amount of power dissipation and significant threshold loss problem make this design less feasible for general operations, especially for the circuits where amount of cascading involved is large. The cascading logic is expected not only to consumes huge amount of power but also due to degraded output logic swings for both sum and carry for a single stage 8T full adder cell, thus the driving capability of this cell is expected to be limited. The methodology that can be used to eradicate this threshold loss problem can be use of a logic implementation technique that itself restores the voltage swings at respective stages or else logic level restoration circuits are used at required places.

Fig 8. 8T Full Adder cell In cases of circuits using this adder cell and a logic level restoration circuit as well, we may be able to eradicate the problem of weak voltage swings but the advantage of using only 8 transistors for implementation of a full adder cell will be lost as the logic level restoring circuit will again increase the over-all area of the circuit.

3. Simulation Results and Analysis We have analyzed all implemented full adder cells for their power , delay and PDP performance using Cadence Virtuoso 90nm technology file and the results are as below with their respective implemented circuits given in the figures from figure 1 to figure 8 . The size of the MOS transistors used is kept as least as possible to support the low area implementation and for any modifications same is reflected in the figures. Table

1: Carry output voltages obtained for respective input combinations (abc)CY

28T

24T

20T

16T

14T

12T

10T

8T

(000)0

.01473E-3

.01468E-3

.008511E-3

.01097E-3

.009866E-3

316.861E-3

262.700E-3

.000156E-3

(001)0

.01673E-3

.015209E-3

.74897E-3

.75534E-3

35.260E-3

318.420E-3

480.800E-3

492.450E-3

(010)0

.01385E-3

.023990E-3

.73686E-3

.008906E-3

.008820E-3

330.760E-3

285.700E-3

.002237E-3

(011)1

1.798E+0

1.7980E+0

1.8000E+0

1.7980E+0

1.7980E+0

1.643E+0

1.79E+0

1.66E+0

(100)0

.01523E-3

.011296E-3

.008081E-3

.74399E-3

.75619E-3

275.170E-3

289.100E-3

.7256E-3

(101)1

1.798E+0

1.7980E+0

1.7980E+0

1.7990E+0

1.7990E+0

1.594E+0

1.8E+0

1.872E+0

(110)1

1.798E+0

1.7980E+0

1.7980E+0

1.7980E+0

1.7980E+0

1.537E+0

1.79E+0

1.799E+0

(111)1

1.798E+0

1.7980E+0

1.8000E+0

1.7990E+0

1.8000E+0

1.594E+0

1.8E+0

1.799E+0

Table 2: Sum output voltages obtained for respective input combinations (abc)SM

28T

24T

20T

16T

14T

12T

10T

8T

000)0

0.0166E-3

0.0175E-3

0.7442E-3

0.8612E-3

0.8097E-3

317.7E-3

277.7E-3

0.0295E-3

(001)1

1.798E+0

1.798E+0

1.798E+0

1.799E+0

1.799E+0

1.64E+0

1.798E+0

1.615E+0

(010)1

1.798E+0

1.789E+0

1.794E+0

1.789E+0

1.791E+0

1.53E+0

1.794E+0

1.421E+0

(011)0

0.0147E-3

0.0147E-3

0.0014E-3

1.525E-3

1.456E-3

0.028E-3

787.2E-3

341.6E-3

(100)1

1.798E+0

1.798E+0

1.794E+0

1.793E+0

1.792E+0

1.798E+0

1.794E+0

1.564E+0

(101)0

0.0201E-3

0.0152E-3

1.468E-3

1.299E-3

1.291E-3

316.6E-3

0.786E-3

267.8E-3

(110)0

0.0170E-3

0.0161E-3

0.7552E-3

0.7454E-3

0.6096E-3

285.2E-3

279.0E-3

136.9E-3

(111)1

1.798E+0

1.798E+0

1.798E+0

1.799E+0

1.799E+0

1.595E+0

1.799E+0

1.609E+0

[469]

Khan et al/COMMUNE – 2015

The inputs a,b and c are provided by the user and CY and SM denoted the expected carry and sum logic levels respectively, where as each row contains the logic voltage levels generated by the adder cell with respective input combination in each row. The ideal voltage generated by an adder cell for logic 0 must be 0V and for logic 1 it should be V dd of the respective circuit or else the maximum input voltage value. Table 1 and 2 show the logic voltages generated for each combinations of the inputs possible starting from (abc) = (000) and analyses it upto all inputs being one i.e (abc) = (111). The best case being that of a 28T adder cell which show very less deviation from the expected voltage levels whereas other cells trade off this deviation of logic voltage levels but at the same time save some power or else some area. Moreover the Bar graphs give a clear idea about the amount of power and respective reduction in number of transistors.

Fig. 9: Delay Analysis

Fig. 10: Power Analysis

Fig. 11: PDP Analysis

4. Conclusion and Future Scope Among various other parameters important for the realization of a VLSI circuit, two critical parameters include power and area. In this work, we have focused on optimization of a base unit of a large system to improve the over-all system without modifying its architecture. The base cell chosen here is the Full Adder Cell. However an important concern is to ensure that the logic voltages

[470]

Khan et al/COMMUNE – 2015

generated at the Sum and the Carry outputs must be such that they are able to drive the subsequent stages attached with these terminals, as a result driving capability becomes a check note for any Full Adder Cell. Hence, this work is an effort towards finding an Adder Cell that can be marked as a reliable adder cell among the implemented cell designs. All the designs have been compared for their power dissipation, delay, and PDP, to choose the more reliable adder cell especially for cascaded VLSI applications. More over tables give the idea about the voltage value associated with both sum and carry outputs for each input combination. After detailed analysis it is observed that 8T transistor Adder cell uses minimum number of transistors in its implementation but has a serious problem of consuming huge amount of power. The huge amount of power consumption of this cell can be accounted to poor Vdd and Gnd isolation and also to use of high power consuming 3-T Xor cell. It was observed that while analyzing the individual power of each MOS cell used to implement the 8-T adder cell, the MOS cells comprising the Xor operation consume maximum power among all other MOS cells. Thus the next best choice for a designer is the 10T adder cell which not only gives the lower amount of the power consumption but also reduces the number of transistors to a great deal. A very close contender to this cell is the 12T adder cell as it gives the lowest amount of power dissipation among all other designs in this work;however, it fails at the check point of having poor driving capabilities, as the amount of degradation in this adder cell is expected to be very large. Thus keeping in view the future VLSI application the preferred choice of a designer must be the 10T adder cell as it not only promises to be low power and low area cell but is also expected to have some significant driving capabilities. This feature of 10T adder cell can prove very useful in systems where the primary job of an adder cell would be to drive another cell. The future scope of this work can be use of this 10T adder cell for some real time VLSI applications such as multipliers, multi-bits adders etc.

5. Acknowledgements We are thankful to the staff of LPU both teaching and non-teaching who stood by our side in terms of providing the required facilities to accomplish our goals.

References R. Zimmermann and W. Fichtner, “Low-power Logic Styles: CMOS versus Pass-Transistor Logic,” IEEE J. Solid-State Circuits, vol. 32, pp.1079-90, July1997. S. Goel, A. Kumar, M. A.Bayoumi, “Design of robust, energy-efficient full adders for deep submicrometer design using hybrid-cmos logic style,” IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 14, pp. 1309–1321, Issue 12, Dec. 2006. N. Weste and K. Eshraghian Principles of CMOS VLSI design, a system perspective, Addison-Wesley, 1993. N. Zhuang and H. Hu, "A new design of the CMOS full adder," IEEE J. of Solid-State Circuits, vol. 21, no. 5, pp. 840-844, May 1992. Chip-Hong Chang, Mingyan Zhang and JiangminGU,”A Novel Low Power Low Voltage Full Adder Cell”, Proceedings of the 3rd International Symposium on Image and Signal Processing and Analysis,Proc. ISPAO3, pp.454-458, 2003. T.Sharma, K.G.Sharma, B.P.Singh, N.Arora, “High Speed, Low Power 8T Full Adder Cell with 45% Improvement in Threshold Loss Problem, “Proceedings of the 12th International Conference on Networking, VLSI and Signal Processing, p. 272, Coimbatore and University of Cambridge, UK, Feb. 2010. Yingtao Jiang, Abdulkarim Al-Sheraidah, Yuke Wang,Edwin Sha and Jin-Gyun Chung, “A Novel Multiplexer –Based Low Power Full Adder”, IEEE Transaction On Circuits And System-II: Express Briefs, Vol. 51 ,No. 7, July 2004. Fayed, AA.; Bayoumi, M.A., "A Low Power 10-Transistor Full Adder Cell for Embedded Architectures," in Proc. IEEE In/. Symp. Circuits and Systems, vol. 4, pp. 226-229, Sydney, Australia, May 2001. Shiwani Singh,Tripti Sharma, K. G. Sharma and Prof. B. P. Singh, “New Design Of Low Power 3t Xor Cell” International Journal Of Computer Engineering &Technology. Vol. 3, Issue 1, pp. 76-80, 2012. Afshan Amin,Shivendra Pandey and Jyotirmoy Pathak,“A Review Paper On 3-T Xor Cells And 8-T Adder Design in Cadence 180nm,”IEEE I2CT,Pune,978-1-4799-3759-2/14,2014.

[471]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Synthesis and Characterization of Chemical Bath Deposited CuZnSnS Nano/Microstructures Suresh Kumar*, Virender Kundu, Mamta, Nikhil Chauhan Department of Electronic Science, Kurukshetra University, Kurukshetra, India.

Abstract In this paper CuZnSnS (CZTS) nano/ microstructure grown on commercial glass slide by chemical bath deposition technique has been presented. The as-deposited CuZnSnS nano/microstructures were characterized by X-ray diffractometer, scanning electron microscope and UV-Vis spectrophotometer. The SEM micrograph confirmed the formation of CuZnSnS nano/microstructures and have ball like structures. The XRD studies confirmed the polycrystalline form consisting of kesterite crystal structures nature of Cu2ZnSnS4 nano/microstructure. The optical energy band gap for CuZnSnS Microstructure was found to be 2.36eV. The CZTS nanostructures have potential application in CZTS nanostructured solar cell.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Chemical Bath Deposition, CZTS, Nano/Microstructures, Photovoltaics

1. Introduction Cu2ZnSnS4 (CZTS) is an important and promising photovoltaic material and has been studied widely for solar cell and other optoelectronic device applications. It is considered an excellent absorber layer in thin film solar cell because of low cost, high absorption coefficient in visible light region, optical band gap of 1.45eV and environmental friendly material (Wang et al, 2010; Katagiri et al, 2008). Many researchers studied and described the preparation of CZTS thin film for their potential application in thin film photovoltaic cells (Katagiri et al, 2008; Barkhouse et al, 2012; Bag et al, 2012; Schurr et al, 2012; Schbert et al, 2011; Weber et al, 2009). (Lin et al, 2012) fabricated nanoparticles based solar cell by using CZTS thin film deposited solution-based chemical route as an absorber layer. (Xosrovashvili, Gorji, 2013) analysed the nanostructured hetro-junction solar cell using TiO2 nanoparticles and CZTS thin film. (Jimbo et al, 2007), (Guo et al, 2009), (Moholkar et al, 2011) and (Dhakal et al, 2014) fabricated CZTS thin film solar cells using costly methods such as electron beam evaporation, hot infusion method, pulsed laser deposition and sputtering respectively. In this paper, CZTS nanostructures have been synthesized by low cost chemical bath deposition technique which can be used for fabricating low cost solar cell.

2. Experimental Details CZTS nanostructures have been deposited by chemical bath deposition technique on commercially available glass slides. The cleaned glass substrate is immersed in dilute solutions containing metals, hydroxide, sulfide ions. The bath was prepared by using 0.2M copper sulfate (CuSO4) in 10ml, 0.2M zinc sulfate (ZnSO4) in 10ml, 0.1M tin chloride(SnCl2)) in 10ml, 0.5M sodium thiosulfate (Na2S2O3.2H2O) in 10ml, 0.1M tri-sodium citrate(C6H5Na3O7.2H2O) in 10ml, 2 ml tri-ethanolamine (TEA) and 5ml ammonia in a beaker. In a beaker, a total volume of 57 ml bath solution was prepared by the sequential addition of all the chemicals. First, the chemicals are added and a solution is prepared prior to the addition of TEA and NH3. The color of the solution prepared was light white color (cream color). Now, TEA was added to this solution which turns the color into dark color. Lastly, NH3 was * Corresponding author. Tel.: +91 9416 377282. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Kumar et al/COMMUNE – 2015

added drop wise and the color of the solution was turned to dark brown and continuously controlled the pH of the final solution to 9.5. Continuous stirring was applied during the whole steps involved in preparing final solution. Then, the cleaned glass slide was immersed vertically in the final solution for 12 hours to grow the CZTS nanostructures at room temperature. The glass slides were taken out from the beaker and dried. The dried sample was then characterized to study its structural, morphological and optical properties. 3. Results and Discussion 3.1

SEM Analysis

The morphological studies of chemical bath deposited CZTS nano/microstructures have been studied by scanning electron microscope. The as-prepared samples were viewed under scanning electron microscope (JEOL JSM-6100) at an accelerating voltage of 15kV under high vacuum. The figures 1a and 1b shows the SEM micrographs of as prepared CuZnSnS (CZTS) nano/microstructures at low and high magnifications. The SEM micrographs show the ball like morphologies of CuZnSnS nano/microstructures. The structures are composite of nano/micro crystals in the size range of nano and micrometers.

Fig. 1 CZTS nano/microstructure (a) low magnification (b) high magnification

3.2

Structural Analysis

XRD is the most essential tool used to characterize crystalline nature of a material. The XRD studies of CuZnSnS nano/microstructures were performed using X’PERT-PRO Phillips X-ray diffractometer using Cu-Kα radiation at 45mA, 45keV. Fig.2 shows the XRD patterns of as-deposited CuZnSnS nano/microstructure. The XRD peaks of (110), (112), (200), (105), (220), (312) and (008) which belong to kesterite CZTS (CZTS JCPDS 26-0575) as observed in the as-prepared sample which indicated the formation of kesterite phase of CZTS structures. The crystalline size of the CZTS was calculated from the Debye-Scherrer’s relation (Alexander, Klug, 1950) and was calculated in the range from 80nm to 1μm.

Fig. 2. XRD pattern of CZTS nano/microstructure

3.3

Optical Analysis

The optical analysis plays an important role in studying the optical properties of a material. The optical absorption properties of CZTS nano/microstructures were studied by Shimadzu 2550 UV-Vis spectrometer in the UV/VIS regions of the electromagnetic spectrum. Fig.3a shows the absorbance spectra of as-deposited CuZnSnS nano/microstructure [473]

Kumar et al/COMMUNE – 2015

 h  vs h ).The optical band gap energy of CZTS nano/microstructures is and Fig.3b shows the Tauc plot of calculated from the plot of the absorption coefficient as a function of wavelength by using Tauc relation (Davis, Mott, 1970) given by 1/ 2

 h 

 h 

1/ 2

Fig. 3 (a) Absorption spectra of CZTS (b) Plot of

vs

n

h

 B  h  Eg 

(1)

for CZTS nano/microstructures

where ‘n’ is the index having values 2, 3, 1/2 and 1/3 corresponding to indirect, indirect forbidden, direct and direct forbidden type of band to band transitions respectively, α is the absorption coefficient, B is a constant called band tailing parameter, h is the Planck’s constant, ν is the frequency of incident radiation photon energy and Eg is the optical energy band gap. The extrapolating of the linear portion of Tauc Plot for n=2 in the equation (1), gives the best linear fit for direct transition. From the fig.3b, it was observed that the optical band gap energy of CZTS nano/microstructures is 2.36eV. The optical band gap energy of CZTS thin film was observed between 1.4-1.6eV as reported by (Lin et al, 2012). It is concluded that the energy band gap of CZTS nano/microstructures is increased as compared to CZTS thin films. 4. Conclusions CZTS nano/microstructures have been prepared on the glass substrates via chemical bath deposition approach at room temperature. The X-ray diffraction study revealed that CuZnSnS nano/microstructures were polycrystalline in nature with kesterite phase. The SEM studies revealed that morphologies of CuZnSnS have ball like nano/microstructures. The optical band gap energy of CuZnSnS nano/microstructure is found to be about 2.36eV. The optical studies revealed that CuZnSnS quartnary materials have wide optical bandgap and thus can be used in optoelectronics and future nano-photovolatics devices. It is concluded that chemical bath deposition method is a simple and useful method for the deposition of CuZnSnS thin film and nano/microstructures. 5. Acknowledgements This work is supported by Science and Engineering Research Board, Department of Science & Technology (DST), Govt. of India (Grant No. SERB/F/2139/2013-14). References Wang K., Gunawan O., Todorov T., Shin B., Chey S.J., Bojarczuk N.A., Mitzi D. and Guha S., 2010, Appl. Phys. Lett. 97, p143508. Katagiri H., Jimbo K., Yamada S., Kamimura T., Maw W.S., Fukano T., Ito T. and Motohiro T., 2008, Appl. Phys. Express 1, p41201. Barkhouse D.A.R., Gunawan O., Gokmen T., Todorov T.K. and D. B. Mitzi,2012, Progr. Photovolt.: Res Appl 20, p6. Bag S., Gunawan O., Gokmen T., Zhu Y., Todorov T.K. and Mitzi D.B., 2012, Energy Environ. Sci. 5, p 7060. Schurr R., Holzing A., Jost S., Hock R., Vo T., Schulze J., Ennaoui A., Lux M., Ahmed S., Reuter K.B., Gunawan O., Guo L., Romankiw L.T. and Deligianni H., 2012, Adv. Energy Mater 2, p253. Schubert B.A., Marsen B., Cinque S., Unold T., Klenk R., Schorr R. and Schock H.W., 2011, Progr. Photovolt.: Res. Appl 19, p93. Weber A., Krauth H., Perlt S., Schubert B., Kotschau I., Schorr S. and Schock H.W., 2009, Thin Solid Films 517, p2524. Lin X., Kavalakkatt J., Kornhuber K., Levcenko S., Lux-Steiner M.C. and Ennaoui A., 2012, Thin Solid Films, http://dx.doi.org/10.1016/j.tsf.2012.10.034. Xosrovashvili G. and Gorji N.E., 2013, Journal of Modern Optics, 60 (11), p936. Jimbo K., Kimura R. and Kamimura T., 2007, Thin Solid Films, 515 (15), p 5997. Guo Q., Hillhouse H.W., and Agrawal R., 2009, Journal of the American Chemical Society, 131(33), p. 11672. Moholkar A.V., Shinde S.S., A. R. Babar et al., 2011, Solar Energy, 85(7), p. 1354. Dhakal T.P., Peng C.Y., Tobias R.R., Dasharathy R. and Westgate C.R., 2014, Solar Energy,100, p23. Alexander L. and Klug H.P., 1950, Journal of Applied Physics 21, p137. Davis E.A. and Mott N.F., 1970, Phil. Magn. 22, p903.

[474]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16-18March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Verification using Multimodal Biometric Fusion Saba Mushtaq*, Shoaib Amin Banday, Ajaz Hussain Mir Department of Electronics and Communication Engineering, National Institute of Technoloy Srinagar, India

Abstract Verification using biometrics has offered a wide range of advantages over conventional possession and knowledge based methods. Almost all the biometric modalities have been tested by now but there are various factors that limit their accuracy. This paper presents a multimodal biometric system for verification. We have fused the matching scores for features extracted from iris and handwritten signatures. GLCM features for iris and GLRLM features for signatures have been used. The verification results so obtained are exceptionally good in comparison to both unimodal iris verification and handwritten signature verification systems.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Multimodal Biometrics; Iris Recognition; Handwritten Signature Verification; Texture Recognition.

1. Introduction A large variety of application requires confirming the identity of the individual before providing them access to the application. Some of these applications need to be highly secured from illegitimate access like bank transaction, security systems, surveillance databases, entrance to high security zone areas etc. Even a small amount of inaccuracy can compromise the access. Biometrics has become a prominent and popular technique to provide personal verification. A biometric system simply matches a pre-stored sample of an individual with the current input sample and matches certain feature to ascertain the identity of the individual. A biometric system operates to either identify or verify an individual. In identification a comparison is made between the submitted sample and all N stored samples in the database (1: N) while as in verification in addition to a submitted sample some pin or password are also entered and a 1:1 comparison is made. A single biometric system may fail to extract enough information for verifying an individual so multimodal biometric that is, biometrics that involve more than one biometric modality to obtain improved performance are used. The most important feature of multimodal systems is to collect information from multiple biometric modalities to reduce the error introduced in monomodal systems (Ross et al 2006). Multimodal systems make it difficult for an intruder to copy more than one biometric traits. The main aim of multimodal systems remains to fuse information obtained from biometric samples at different fusion levels (Rattani et al. 2006). This fusion can be performed at four different levels sensor level, feature level, matching level and decision level. The first two levels i.e. sensor and the feature level are referred to as a pre-mapping fusion while as if the matching is performed at matching score level and the decision level then it is referred to as a post-mapping fusion (Sanderson.C, K. K. Paliwal 2003). In this paper, we fuse the information at matching score level. We have made use of GLCM (Grey level co-occurrence matrix) to calculate features of right iris and GLRLM (Grey level rum length matrix) to calculate features of handwritten signature images. Iris recognition is the most promising for high security environments (J. Daugman 1993). Iris based biometric recognition systems have achieved a very high accuracy as high as 97% (C. Sanchez-Avila et al. 2001). A brief description of the two texture based techniques viz GLCM and GLRLM are given in next section.

*

Corresponding author. Tel.: +91 9906 118357. E-mail address: [email protected]. ISBN: 978-93-82288-63-3

Mushtaq et al./ COMMUNE-2015

2. Generalized Description of Feature Extraction schemes Texture is one of the important characteristics used in identifying an object in an image and to discriminate the images. The texture coarseness or fineness of an image can be interpreted as the distribution of the elements in the matrix (Harlic 1973). The gray tone spatial dependence was first used by ( Julesz 1962) for texture classification. 2.1. GLCM GLCM is a second order statistics method, which describes the spatial interrelationships of the gray tones in an image (R.W. Conners and C. A. Harlow, 1980). It contains elements that are counts of the number of pixel pairs, which are separated by certain distance and at some angular direction. Typically, GLCM is calculated in a small window, which scans the whole image. Bachoo and Tapamo (2005) have used GLCM for pattern analysis of iris however in this method the selection of window size remains a problem. In the proposed scheme we have normalize, the GLCM and assumed GLCM represent probabilities instead of counts. The co-occurrence matrix is constructed by the joint probability density function between the gray level tones, which gives the spatial relationship between any two points in the image. It is denoted by P(i,j,d,θ), where i and j give ith line and jth column of co-occurrence matrix respectively, d is the distance between any two points and θ is the direction. Normalization involves dividing by the total number of counted pixel pairs. There are eight texture features based on GLCM as studied by (Haralick 1973). The mathematical expressions for these features are given below:

Entropy

=

(1)

Correlation

=

(2)

Contrast

=

(3)

Dissimilarity

=

(4)

Homogeneity

=

(5)

Angular Second Moment

=

(6)

Mean

=

i(µ)i

(7)

Variance

=

(σ 2 )

(8)

Herei GLCM features are computed based on two parameters, which are the distance between the pixel pair ‘d’ and their angular relation θ. The angular relation is quantized at four angles i.e., 00, 450, 900 and 1800. 2.2. GLRLM The technique used to calculate features of handwritten signatures is GLRLM. The GLRLM is based on computing the number of gray-level runs of various lengths. A gray level run is a set of consecutive and collinear pixel points having the same gray level value. The length of the run is the number of pixel points in the run. The Gray Level Run Length matrix is constructed as follows:

R(θ) = ( g (i,j) | θ ), 0 ≤ I ≤ Ng , 0 ≤ I ≤ Rmax;

(9)

Where Ng is the maximum gray level and Rmax is the maximum length. Let p (i, j) be the number of times there is a run of length j having gray level i. There are five Run Length Matrix based features computed for 4 directions of run (0°, 45°, 90°, 135°). For each matrix in a particular direction following seven GLRLM features viz SRE, LRE, GLN, RLN, RP, LGLRE, HGLRE are obtained. These features were suggested by Gallow 1975

[476]

Mushtaq et al./COMMUNE – 2015 Table 1 : GLRLM features

S.NO.

Features

1

Short Run Emphasis(SRE)

2

Long Run Emphasis(LRE)

3

Grey Level Nonuniformity(GLN)

4

Run Length NonUniformity

5

Run Percentage(RP)

6

Low Grey level Run Emphasis(LGLRE)

7

High Grey Level Run Emphasis(HGLRE)

Formulae

3. Proposed Scheme This section presents the proposed scheme for verification using multimodal biometric fusion. 3.1. Database We have used two databases one for signatures and one for iris images. For iris experiments are carried on CASIAiris-V4 thousand database and for hand written signatures a signature database collected at NIT Srinagar is used. 3.2. Algorithm The above discussed textural feature extraction method are used to extract features from right iris and hand written signatures. GLCM textural features are calculated using equations shown in section 2. The distance d between pixel pairs is first selected as 1 and gray level co-occurrence matrix features are calculated. The features are calculated at an angles of 00,450,900 and 1800. The experiment is continued with calculation of GLCM features for value of d=2 and 4. The features are again measured at the pre-defined angles of rotation over the whole iris image. To make the GLCM invariant to the rotation of the images, GLCM obtained at d = 1,2 and 4 is averaged through four angular relations (00,450,900 and 1800 ). Once the textural features of image (iris) is calculated, the feature set is stored in the database as trainer. The test image (iris) is similarly processed to obtain a textural feature vector. This feature vector of the test image is processed by the matching unit of the multi unit biometric verification system and compared against the templates stored in the database. The matching unit because of taking Euclidean distance as the classifier outputs a matching score corresponding to each template in the database. The matching score is fed to the decision unit which because of some predefined threshold classifies test image as Genuine or Imposter based on the score obtained at the matching unit. Similarly, the signature image is pre- processed separately as shown in the figure 1. The GLRLM features are extracted for the signature image in the similar way followed by matching. In case Once the scores are obtained from both the matching units ( i-e., from right iris and handwritten signatures), they are fused using SUM method of fusion.

[477]

Mushtaq et al./ COMMUNE-2015

Fig 1: Block Diagram of proposed system

4. Experimental Results We evaluated the proposed system on a data set of 525 signature images collected at NIT Srinagar and CASIA-irisV4 thousand database. The training set contains 200 each, a signature and an iris image assumed to be belonging to same individual. The testing set is a set of 50 each, signatures and iris images for verification. 4.1

Training

For an individual one image for each trait i.e iris and signature is enrolled in the database for which features are extracted using GLCM and GLRLM respectively. Which are also used for score level fusion and saved in database for verification. 4.2

Testing

Pair of iris and signature are used for testing.. Fused feature vector is generated from the pair and is compared with the database score value.We have calculated FAR and Accuracy for the individual systems and then for the multimodal proposed system. The results thus obtained are given in table below. We also calculated the Genuine and imposter scores for the two systems that is for iris using GLCM and for Signatures using GLRLM and for the proposed system which are given below in fig2, fig 3 and fig4 respectively . It can be clearly seen that the overlap of genuine and imposter scores is more in unimodal systems and for every verification systems the aim is to reduce this overlap as much as possible. Lesser the overlap more accurate is the system.

[478]

Mushtaq et al./COMMUNE – 2015

Fig 2 : Score distribution for iris using GLCM

Fig 3 : Score distribution for Signature using GLRLM

Fig 4 Score density for proposed system. Table 2: Accuracy and FAR S. No.

Modality

FAR

Accuracy

1 2 3

Signatures Iris Proposed Multimodal system

13.33 0.8 0.17

85.15% 93.2% 97.6%

5. Conclusion The field of multimodal biometrics is a challenging and novel area of research aiming at reliable and accurate personal verification and identification. This paper presents a score level fusion technique for human verification. The proposed technique uses texture features viz GLCM and GLRLM for feature extraction. The features thus obtained are normalized. In addition, stored in database. Matching scores are calculated for iris and signatures separately, then fused using sum rule of fusion, and results are compared to unimodal systems based on accuracy and FAR. The experimental results establish the effectiveness of fusion of the individual matching scores and accuracy of 97.6% is abtained in comparision to individual signature and iris biometric systems that provide accuracy of 85.15% and 93.2 % respectively. References Asheer Kasar Bachoo and Jules-Raymond Tapamo : Texture detection for segmentation of iris images. In Annual Research Conference of the South African Institute of Computer Information Technologists, pages 236–243, 2005. C. Sanchez-Avila, R. Sanchez-Reil, D. de Martin-Roche : Iris-Based Biometric Recognition using Dyadic Wavelet Transform , IEEE AESS Systems Magazine, October 2002 Ross. A., Nandakumar, K., Jain, A.K.: Handbook of Multibiometrics. Springer Verlag (2006) Rattani, A., Kisku, D.R., Bicego, M., Tistarelli, M.: Robust Feature-Level Multibiometrics Classification. IEEE Biometric Consortium Conference, Biometrics Symposium, pp. 1—6 (2006) R.W. Conners and C. A. Harlow. A theoretical comparison of texture algorithms. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2(3):204–222, 1980. Sanderson.C, K. K. Paliwal. Information Fusion and Person Verification Using Speech and Face Information. IDIAP-RR, pp.02-33, 2003. Haralick, R.M. , Shanmugan, K.. and Dinstein, I.( 1973) ‘Textural Features for Image Classification’, IEEE Tr. on Systems, Man, and Cybernetics, Vol SMC-3, No. 6, pp. 610-621 J. Daugman, November 1993, High confidence visual recognition by test of statistical independence, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 15, pp. 1148-1 161. Julesz B, “Visual pattern discrimination,” IRE T M In~form, Theory, vol. 8, no. 2, pp. 84-92,Feb. 1962. M. M. Galloway,“Texture analysis using gray level run lengths”, Computer Graphics Image Process., Vol. 4, pp. 172–179, June 1975.

[479]

2015 International Conference on Advances in

Computers, Communication and Electronic Engineering 16 -18 March, 2015

PG Department of Electronics and Instrumentation Technology University of Kashmir, Srinagar, India

Extension to the K-Means Algorithm for Automatic Generation of Clusters for Mixed Datasets Anupama Chadhaa*, Suresh Kumarb a Faculty of Computer Applications, MRIU, Faridabad, India Faculty of Engineering and Technology, MRIU, Faridabad, India

*

Abstract A lot of work has been done and is still in progress on the famous partition based K-Means clustering algorithm. Various forms of KMeans have been proposed depending on the type of data sets being handled. Most popular ones are K-Modes for categorical data and K-Prototype for mixed numerical and categorical data. In all these forms of K-Means, one major limitation is dependency on prior input of number of clusters K, and sometimes it becomes practically impossible to correctly estimate the optimum number of clusters in advance. Various ways have been suggested in literature to overcome this limitation for numerical data. But for categorical and mixed data work is still in progress. In this paper, we have proposed a method for clustering mixed data, which is based on K-Means, but has advanced features for automatic generation of appropriate number of clusters.

© 2015 Published by University of Kashmir, Srinagar. Selection and/or peer-review under responsibility of Department of Electronics and Instrumentation Technology, University of Kashmir, Srinagar. Keywords: Clustering; K-Means; Mixed Dataset; Automatic Generation of Clusters

1. Introduction Clustering is the method of segregating the objects into groups such that the objects in a certain cluster have high degree of similarity with each other than the objects in the other clusters. The clusters formed are then analyzed for making decisions . For example, in Educational Institutes, clustering students according to their academic performance can help in identifying the weaker ones, who can be provided with more tuitions. In Health sector, data mining and clustering may help in identifying disease symptoms, ethnicities, water quality etc. and establishing links between them. In Banking sector it can be used to group customers with overdue credit card payments. In Market Research, Data Mining and Clustering can be used to identify customers having certain buying patterns. Number of clustering methods have been suggested in the literature (Guojun et al, 2007; RuiXu et al, 2005). Out of these, the partition based method is known for its speed in clustering large data sets. One of the famous algorithm based on this method is K-Means. K-Means is a simple algorithm known for its speed. However, there exist some limitations in this algorithm. One major limitation is the requirement to input the number of clusters at the very beginning based on anticipation.. This is domain specific, and if the person using the algorithm is not domain expert, then an incorrect number of clusters may be input, leading to inefficient grouping. Work has been done for auto generation of clusters with numerical data. However, for categorical and mixed type of data, this limitation still exists. In this paper we have proposed an algorithm based on K-Protoype, which is an extension of K-Means to deal with mixed type of data to automatically generate suitable number of clusters. 2. Literature Survey 2.1. This part of the literature survey discusses the works of the authors done to remove the limitation of inputting the number of clusters required with numerical data set in K-Means.

*

Corresponding Author. Tel.: +91 9818 856542. Email-Address: [email protected]. ISBN: 978-93-82288-63-3

Chadha and Kumar/COMMUNE – 2015

Pelleg et al. (2000) suggested XMeans algorithm as an extension of K-Means which required the user to input a range representing the lower and upper instead of a particular value of K. The algorithm initially takes lower bound of the given range as K and continues to add centroids until the upper bound is reached. The algorithm terminates as soon as it gets the centroid set that scores the best. The drawback lies in the fact that it requires the user to input a range suggesting the lower and upper bound of K. Tibshirani et al. (2000) used the technique of Gap Statistic. In this technique output generated by any clustering algorithm was used to compare the change in within cluster dispersion to that expected under an appropriate reference null distribution. The algorithm works well with well separated clusters. Wagstaff et al. (2001) suggested utilizing information about the problem domain in order to put some constraints on the data set. During the clustering process it is ensured that none of the constraint is violated. This algorithm requires some domain specific information, which sometimes becomes difficult to attain. Cheung (2003) proposed a new extension of K-Means clustering technique named STep-wise Automatic Rival penalized (STAR) K-Means algorithm overcoming two of its major limitations of dependency on initial centroids and inputting K. In the first step of the algorithm cluster centres are provided and in the second step the units are adjusted adaptively by a learning rule. The limitation of this algorithm is the complex computation involved in it. Shafeeq et al. (2012) proposed an algorithm in which the optimal number of clusters was found on the run. The main drawback of the proposed approach is that its computational time is more than the K-Means for larger data sets. Also the user has to input the number of clusters (K) as 2 in the first run. Leela et al. (2013) proposed Y-means algorithm. Initially, clusters are found using K-Means algorithm on the data set. A sequence of splitting, deleting and merging the clusters is then followed to find the optimal number of clusters. The limitation of this algorithm is that it depends on K-Means algorithm to find the clusters initially. Abubaker et al. (2013) presented a new approach to based on the K-Nearest Neighbour method. The only input parameter taken by the algorithm is kn (the number of nearest neighbour). The drawback of this algorithm is that we have to input the number of nearest neighbours kn. 2.2. In this part, the work done in the field of automatic generation of clusters with categorical data has been discussed. Not much of the work has been done in this field. In this section we will be discussing two of the research papers dealing with this limitation of inputting the value of K. Liao et al. (2009) proposed the new algorithm which extends the K-Modes clustering algorithm by introducing a regularization parameter to control the number of clusters in a clustering process. A suitable value of regularization parameter is chosen to generate the most stable clustering result. The major limitation of the above proposed algorithm is that the computation involved is much more than the original K-Modes algorithm. Also this algorithm requires an input parameter representing the initial cluster centres. Cheung et al. (2013) proposed a similarity metric that can be applied to categorical, numerical, and mixed attributes. Based on this similarity metric an iterative clustering algorithm is developed to overcome the limitation of inputting K. This algorithm requires some initial value of K which should not be less than the original value of K. The cluster accuracy is more as compared to the original K-Modes and K-Modes with Ng’s dissimilarity metric (Ng, 2007). But as in the clustering algorithm proposed by (Liao, 2009), this algorithm too has much computation involved in it. 2.3. Very little work has been done to automatically generate clusters in a dataset containing mixed attributes. Two research papers providing solution to this problem are discussed below: Liang et al. (2012) extended K-Prototype algorithm by proposing a new dissimilarity measure for mixed data set. A mechanism was developed to make use of within-cluster entropy and between-cluster entropy to identify the worst cluster in a mixed dataset. The major limitation of the above proposed algorithm is that this algorithm requires input parameters representing the minimum and maximum number of clusters that can be generated from the data set. Ahmad et al. (2007) proposed a new cost function for mixed data set. The authors extended K-Means algorithm that worked well for data with mixed numeric and categorical features by introducing a new distance measure and a new way of finding the centroids of the clusters. The algorithm used the concept of the significance of an attribute in the clustering process. In this algorithm computation times increases with the increase in the dimensions of the data. 3. Proposed Algorithm As discussed in section 2.3, the algorithms proposed to overcome the limitation of inputting the value of K for mixed data require either some input parameter or some initial value of K to achieve good clustering results. In the proposed algorithm we have extended K-Prototype algorithm to overcome this limitation. In the proposed algorithm we have utilized the methods proposed by (Ahmad et al, 2007) to find the centroids of the clusters and the distance of the records from the centroid. Also we have utilized the concept of the most significant attribute suggested by (Ahmad et al, 2007) to create initial clusters. To compute the significance of numeric attributes, they have been discretized using equal

[481]

Chadha and Kumar/COMMUNE – 2015

width interval method (Ahmad et al, 2007). Input: dataset of n objects with m mixed attributes. Output: clusters or groups distributing the objects in the given dataset. a) b) c) d)

e)

f) g) h) i) j) k)

Discretize all numerical attributes and find the most significant attribute using the way proposed by (Ahmad et al, 2007) Create initial clusters with attribute values with maximum distance in most significant attribute in different clusters. Normalize all the numerical attributes and find the centroids of all the clusters using the way proposed by (Ahmad et al, 2007). Find the distance of every tuple from the respective centroids using the way proposed by (Ahmad et al, 2007). The minimum of all the average distance of every cluster of all the distance values is taken as d. Find the outliers in the initial two clusters according to the following objective function: i. Distance(x, q)