Computational intelligence in data mining (vol 2) jain, behera, mandal mohapatra 2014 12 11

Saibaba Reddy Vice Chancellor, VSSUT, Burla, Odisha, India Honorary General Chair Prof.. Rajib Mall, Ph.D., Professor and Head Department of Computer Science and Engineering, IIT Kharagp

Lakhmi C Jain

Himansu Sekhar Behera

Jyotsna Kumar Mandal

Durga Prasad Mohapatra

Volume 32

Series editors

Robert J Howlett, KES International, Shoreham-by-Sea, UK

e-mail: rjhowlett@kesinternational.org

Lakhmi C Jain, University of Canberra, Canberra, Australia, and

University of South Australia, Adelaide, Australia

e-mail: Lakhmi.jain@unisa.edu.au

The Smart Innovation, Systems and Technologies book series encompasses thetopics of knowledge, intelligence, innovation and sustainability The aim of theseries is to make available a platform for the publication of books on all aspects ofsingle and multi-disciplinary research on these themes in order to make the latestresults available in a readily-accessible form Volumes on interdisciplinaryresearch combining two or more of these areas is particularly sought.

The series covers systems and paradigms that employ knowledge andintelligence in a broad sense Its scope is systems having embedded knowledgeand intelligence, which may be applied to the solution of world problems inindustry, the environment and the community It also focusses on the knowledge-transfer methodologies and innovation strategies employed to make this happeneffectively The combination of intelligent systems tools and a broad range ofapplications introduces a need for a synergy of disciplines from science,technology, business and the humanities The series will include conferenceproceedings, edited collections, monographs, handbooks, reference books, andother relevant types of book in areas of science and technology where smartsystems and technologies can offer innovative solutions

High quality content is an essential feature for all book proposals accepted forthe series It is expected that editors of all accepted volumes will ensure thatcontributions are subjected to an appropriate level of reviewing process and adhere

to KES quality principles

More information about this series at http://www.springer.com/series/8767

Lakhmi C Jain Himansu Sekhar Behera

Durga Prasad Mohapatra

Editors

Computational Intelligence

in Data Mining - Volume 2

Proceedings of the International Conference

on CIDM, 20-21 December 2014

123

Himansu Sekhar Behera

Department of Computer Science

Durga Prasad MohapatraDepartment of Computer Scienceand Engineering

National Institute of Technology RourkelaRourkela

India

ISSN 2190-3018 ISSN 2190-3026 (electronic)

Smart Innovation, Systems and Technologies

ISBN 978-81-322-2207-1 ISBN 978-81-322-2208-8 (eBook)

DOI 10.1007/978-81-322-2208-8

Library of Congress Control Number: 2014956493

Springer New Delhi Heidelberg New York Dordrecht London

© Springer India 2015

This work is subject to copyright All rights are reserved by the Publisher, whether the whole or part

of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission

or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed.

The use of general descriptive names, registered names, trademarks, service marks, etc in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.

The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made.

Printed on acid-free paper

Springer (India) Pvt Ltd is part of Springer Science+Business Media (www.springer.com)

The First International Conference on“Computational Intelligence in Data Mining(ICCIDM-2014)” was hosted and organized jointly by the Department of ComputerScience and Engineering, Information Technology and MCA, Veer Surendra SaiUniversity of Technology, Burla, Sambalpur, Odisha, India between 20 and 21December 2014 ICCIDM is an international interdisciplinary conference coveringresearch and developments in the fields of Data Mining, Computational Intelli-gence, Soft Computing, Machine Learning, Fuzzy Logic, and a lot more More than

550 prospective authors had submitted their research papers to the conference.ICCIDM selected 192 papers after a double blind peer review process by experi-enced subject expertise reviewers chosen from the country and abroad Theproceedings of ICCIDM is a nice collection of interdisciplinary papers concerned invarious prolific research areas of Data Mining and Computational Intelligence Ithas been an honor for us to have the chance to edit the proceedings We haveenjoyed considerably working in cooperation with the International Advisory,Program, and Technical Committees to call for papers, review papers, andfinalizepapers to be included in the proceedings

This International Conference ICCIDM aims at encompassing a new breed ofengineers, technologists making it a crest of global success It will also educate theyouth to move ahead for inventing something that will lead to great success Thisyear’s program includes an exciting collection of contributions resulting from asuccessful call for papers The selected papers have been divided into thematicareas including both review and research papers which highlight the current focus

of Computational Intelligence Techniques in Data Mining The conference aims atcreating a forum for further discussion for an integrated information field incor-porating a series of technical issues in the frontier analysis and design aspects ofdifferent alliances in the relatedfield of Intelligent computing and others Thereforethe call for paper was on three major themes like Methods, Algorithms, and Models

in Data mining and Machine learning, Advance Computing and Applications.Further, papers discussing the issues and applications related to the theme of theconference were also welcomed at ICCIDM

v

The proceedings of ICCIDM have been released to mark this great day inICCIDM which is a collection of ideas and perspectives on different issues andsome new thoughts on various fields of Intelligent Computing We hope theauthor’s own research and opinions add value to it First and foremost are theauthors of papers, columns, and editorials whose works have made the conference agreat success We had a great time putting together this proceedings The ICCIDMconference and proceedings are a credit to a large group of people and everyoneshould be there for the outcome We extend our deep sense of gratitude to all fortheir warm encouragement, inspiration, and continuous support for making itpossible.

Hope all of us will appreciate the good contributions made and justify ourefforts

The theme and relevance of ICCIDM has attracted more than 550 researchers/academicians around the globe, which enabled us to select good quality papers andserve to demonstrate the popularity of the ICCIDM conference for sharing ideasand researchfindings with truly national and international communities Thanks toall who have contributed in producing such a comprehensive conferenceproceedings of ICCIDM.

The organizing committee believes and trusts that we have been true to the spirit

of collegiality that members of ICCIDM value, even as maintaining an elevatedstandard as we have reviewed papers, provided feedback, and present a strong body

of published work in this collection of proceedings Thanks to all the members

of the Organizing committee for their heartfelt support and cooperation

It has been an honor for us to edit the proceedings We have enjoyed erably working in cooperation with the International Advisory, Program, andTechnical Committees to call for papers, review papers, andfinalize papers to beincluded in the proceedings

consid-We express our sincere thanks and obligations to the benign reviewers forsparing their valuable time and effort in reviewing the papers along with sugges-tions and appreciation in improvising the presentation, quality, and content of thisproceedings Without this commitment it would not be possible to have theimportant reviewer status assigned to papers in the proceedings The eminence

of these papers is an accolade to the authors and also to the reviewers who haveguided for indispensable perfection

We would like to gratefully acknowledge the enthusiastic guidance and tinuous support of Prof (Dr.) Lakhmi Jain, as and when it was needed as well asadjudicating on those difficult decisions in the preparation of the proceedings andimpetus to our efforts to publish this proceeding

con-Last but not the least, the editorial members of Springer Publishing deserve aspecial mention and our sincere thanks to them not only for making our dreamcome true in the shape of this proceedings, but also for its brilliant get-up andin-time publication in Smart, Innovation, System and Technologies, Springer

vii

I feel honored to express my deep sense of gratitude to all members of InternationalAdvisory Committee, Technical Committee, Program Committee, OrganizingCommittee, and Editorial Committee members of ICCIDM for their unconditionalsupport and cooperation.

The ICCIDM conference and proceedings are a credit to a large group of peopleand everyone should be proud of the outcome

Himansu Sekhar Behera

The International Conference on “Computational Intelligence in Data Mining”(ICCIDM-2014) has been established itself as one of the leading and prestigiousconference which will facilitate cross-cooperation across the diverse regionalresearch communities within India as well as with other International regionalresearch programs and partners Such an active dialogue and discussion amongInternational and National research communities is required to address many newtrends and challenges and applications of Computational Intelligence in thefield ofScience, Engineering and Technology ICCIDM 2014 is endowed with an oppor-tune forum and a vibrant platform for researchers, academicians, scientists, andpractitioners to share their original research findings and practical developmentexperiences on the new challenges and budding confronting issues.

The conference aims to:

• Provide an insight into current strength and weaknesses of current applications aswell as researchfindings of both Computational Intelligence and Data Mining

• Improve the exchange of ideas and coherence between the various ComputationalIntelligence Methods

• Enhance the relevance and exploitation of data mining application areas for user as well as novice user application

end-• Bridge research with practice that will lead to a fruitful platform for the opment of Computational Intelligence in Data mining for researchers andpractitioners

devel-• Promote novel high quality research findings and innovative solutions to thechallenging problems in Intelligent Computing

• Make a tangible contribution to some innovative findings in the field of datamining

• Provide research recommendations for future assessment reports

ix

So, we hope the participants will gain new perspectives and views on currentresearch topics from leading scientists, researchers, and academicians around theworld, contribute their own ideas on important research topics like Data Mining andComputational Intelligence, as well as network and collaborate with their interna-tional counterparts.

Prof E Saibaba Reddy

Vice Chancellor, VSSUT, Burla, Odisha, India

Honorary General Chair

Prof P.K Dash, Ph.D., D.Sc., FNAE, SMIEEE, Director

Multi Disciplinary Research Center, S‘O’A University, India

Prof Lakhmi C Jain, Ph.D., M.E., B.E.(Hons), Fellow (Engineers Australia),University of Canberra, Canberra, Australia and University of South Australia,Adelaide, SA, Australia

Honorary Advisory Chair

Prof Shankar K Pal, Distinguished Professor

Indian Statistical Institute, Kolkata, India

General Chair

Prof Rajib Mall, Ph.D., Professor and Head

Department of Computer Science and Engineering, IIT Kharagpur, India

xi

Program Chairs

Dr Sukumar Mishra, Ph.D., Professor

Department of EE, IIT Delhi, India

Prof H.S Behera, Reader, Department of Computer Science Engineering andInformation Technology, Veer Surendra Sai University of Technology, Burla,Odisha, India

Prof J.K Mandal, Professor, Department of Computer Science and Engineering,University of Kalyani, Kolkata, India

Prof D.P Mohapatra, Associate Professor, Department of Computer Science andEngineering, NIT, Rourkela, Odisa, India

International Advisory Committee

Prof C.R Tripathy (VC, Sambalpur University)

Prof B.B Pati (VSSUT, Burla)

Prof A.N Nayak (VSSUT, Burla)

Prof S Yordanova (STU, Bulgaria)

Prof P Mohapatra (University of California)

Prof S Naik (University of Waterloo, Canada)

Prof S Bhattacharjee (NIT, Surat)

Prof G Saniel (NIT, Durgapur)

Prof K.K Bharadwaj (JNU, New Delhi)

Prof Richard Le (Latrob University, Australia)

Prof K.K Shukla (IIT, BHU)

Prof G.K Nayak (IIIT, BBSR)

Prof S Sakhya (TU, Nepal)

Prof A.P Mathur (SUTD, Singapore)

Prof P Sanyal (WBUT, Kolkata)

Prof Yew-Soon Ong (NTU, Singapore)

Prof S Mahesan (Japfna University, Srilanka)

Prof B Satapathy (SU, SBP)

Prof G Chakraborty (IPU, Japan)

Prof T.S Teck (NU, Singapore)

Prof P Mitra (P.S University, USA)

Prof A Konar (Jadavpur University)

Prof S Das (Galgotias University)

Prof A Ramanan (UJ, Srilanka)

Prof Sudipta Mohapatra, (IIT, KGP)

Prof P Bhattacharya (NIT, Agaratala)

Prof N Chaki (University of Calcutta)

Dr J.R Mohanty, (Registrar, VSSUT, Burla)

Prof M.N Favorskaya (SibSAU)

Mr D Minz (COF, VSSUT, Burla)

Prof L.M Patnaik (DIAT, Pune)

Prof G Panda (IIT, BHU)

Prof S.K Jena (NIT, RKL)

Prof V.E Balas (University of Arad)

Prof R Kotagiri (University of Melbourne)

Prof B.B Biswal (NIT, RKL)

Prof Amit Das (BESU, Kolkata)

Prof P.K Patra (CET, BBSR)

Prof N.G.P.C Mahalik (California)

Prof D.K Pratihar (IIT, KGP)

Prof A Ghosh (ISI, Kolkata)

Prof P Mitra (IIT, KGP)

Prof P.P Das (IIT, KGP)

Prof M.S Rao (JNTU, HYD)

Prof A Damodaram (JNTU, HYD)

Prof M Dash (NTU, Singapore)

Prof I.S Dhillon (University of Texas)

Prof S Biswas (IIT, Bombay)

Prof S Pattnayak (S‘O’A, BBSR)

Prof M Biswal (IIT, KGP)

Prof Tony Clark (M.S.U, UK)

Prof Sanjib ku Panda (NUS)

Prof G.C Nandy (IIIT, Allahabad)

Prof R.C Hansdah (IISC, Bangalore)

Prof S.K Basu (BHU, India)

Prof P.K Jana (ISM, Dhanbad)

Prof P.P Choudhury (ISI, Kolkata)

Prof H Pattnayak (KIIT, BBSR)

Prof P Srinivasa Rao (AU, Andhra)

International Technical Committee

Dr Istvan Erlich, Ph.D., Chair Professor, Head

Department of EE and IT, University of DUISBURG-ESSEN, Germany

Dr Torbjørn Skramstad, Professor

Department of Computer and System Science, Norwegian University

of Science and Technology, Norway

Dr P.N Suganthan, Ph.D., Associate Professor

School of EEE, NTU, Singapore

Prof Ashok Pradhan, Ph.D., Professor

Department of EE, IIT Kharagpur, India

Head, Department of IT, B.I.T, Meshra, India

Dr Amit Saxena, Ph.D., Professor

Head, Department of CS and IT, CU, Bilashpur, India

Dr Sidhartha Panda, Ph.D., Professor

Department of EEE, VSSUT, Burla, Odisha, India

Dr Swagatam Das, Ph.D., Associate Professor

Indian Statistical Institute, Kolkata, India

Dr Chiranjeev Kumar, Ph.D., Associate Professor and Head

Department of CSE, Indian School of Mines (ISM), Dhanbad

Dr B.K Panigrahi, Ph.D., Associate Professor

Department of EE, IIT Delhi, India

Dr A.K Turuk, Ph.D., Associate Professor

Head, Department of CSE, NIT, RKL, India

Dr S Samantray, Ph.D., Associate Professor

Department of EE, IIT BBSR, Odisha, India

Dr B Biswal, Ph.D., Professor

Department of ETC, GMRIT, A.P., India

Dr Suresh C Satpathy, Professor, Head

Department of Computer Science and Engineering, ANITS, AP, India

Dr S Dehuri, Ph.D., Associate Professor

Department of System Engineering, Ajou University, South Korea

Dr B.B Mishra, Ph.D., Professor,

Department of IT, S.I.T, BBSR, India

Dr G Jena, Ph.D., Professor

Department of CSE, RIT, Berhampu, Odisha, India

Dr Aneesh Krishna, Assistant Professor

Department of Computing, Curtin University, Perth, Australia

Dr Ranjan Behera, Ph.D., Assistant Professor

Department of EE, IIT, Patna, India

Dr A.K Barisal, Ph.D., Reader

Department of EE, VSSUT, Burla, Odisha, India

Dr R Mohanty, Reader

Department of CSE, VSSUT, Burla

Conference Steering Committee

Publicity Chair

Prof A Rath, DRIEMS, Cuttack

Prof B Naik, VSSUT, Burla

Mr Sambit Bakshi, NIT, RKL

Logistic Chair

Prof S.P Sahoo, VSSUT, Burla

Prof S.K Nayak, VSSUT, Burla

Prof D.C Rao, VSSUT, Burla

Prof K.K Sahu, VSSUT, Burla

Organizing Committee

Prof D Mishra, VSSUT, Burla

Prof J Rana, VSSUT, Burla

Prof P.K Pradhan, VSSUT, Burla

Prof P.C Swain, VSSUT, Burla

Prof P.K Modi, VSSUT, Burla

Prof S.K Swain, VSSUT, Burla

Prof P.K Das, VSSUT, Burla

Prof P.R Dash, VSSUT, Burla

Prof P.K Kar, VSSUT, Burla

Prof U.R Jena, VSSUT, Burla

Prof S.S Das, VSSUT, Burla

Prof Sukalyan Dash, VSSUT, Burla

Prof D Mishra, VSSUT, Burla

Prof S Aggrawal, VSSUT, Burla

Prof R.K Sahu, VSSUT, Burla

Prof M Tripathy, VSSUT, Burla

Prof K Sethi, VSSUT, Burla

Prof B.B Mangaraj, VSSUT, Burla

Prof M.R Pradhan, VSSUT, Burla

Prof S.K Sarangi, VSSUT, Burla

Prof N Bhoi, VSSUT, Burla

Prof J.R Mohanty, VSSUT, Burla

Prof Sumanta Panda, VSSUT, Burla

Prof A.K Pattnaik, VSSUT, Burla

Prof S Panigrahi, VSSUT, Burla

Prof S Behera, VSSUT, Burla

Prof M.K Jena, VSSUT, Burla

Prof S Acharya, VSSUT, Burla

Prof S Kissan, VSSUT, Burla

Prof S Sathua, VSSUT, Burla

Prof E Oram, VSSUT, Burla

Dr M.K Patel, VSSUT, Burla

Mr N.K.S Behera, M.Tech Scholar

Mr T Das, M.Tech Scholar

Mr S.R Sahu, M.Tech Scholar

Mr M.K Sahoo, M.Tech Scholar

Prof J.V.R Murthy, JNTU, Kakinada

Prof G.M.V Prasad, B.V.CIT, AP

Prof S Pradhan, UU, BBSR

Prof P.M Khilar, NIT, RKL

Prof Murthy Sharma, BVC, AP

Prof M Patra, BU, Berhampur

Prof M Srivastava, GGU, Bilaspur

Prof P.K Behera, UU, BBSR

Prof B.D Sahu, NIT, RKL

Prof S Baboo, Sambalpur University

Prof Ajit K Nayak, S‘O’A, BBSR

Prof Debahuti Mishra, ITER, BBSR

Prof S Sethi, IGIT, Sarang

Prof C.S Panda, Sambalpur University

Prof N Kamila, CVRCE, BBSR

Prof H.K Tripathy, KIIT, BBSR

Prof S.K Sahana, BIT, Meshra

Prof Lambodar Jena, GEC, BBSR

Prof R.C Balabantaray, IIIT, BBSR

Prof D Gountia, CET, BBSR

Prof Mihir Singh, WBUT, Kolkata

Prof A Khaskalam, GGU, Bilaspur

Prof Sashikala Mishra, ITER, BBSR

Prof D.K Behera, TAT, BBSR

Prof Shruti Mishra, ITER, BBSR

Prof H Das, KIIT, BBSR

Mr Sarat C Nayak, Ph.D Scholar

Mr Pradipta K Das, Ph.D Scholar

Mr G.T Chandrasekhar, Ph.D Scholar

Mr P Mohanty, Ph.D Scholar

Mr Sibarama Panigrahi, Ph.D Scholar

Mr A.K Bhoi, Ph.D Scholar

Mr T.K Samal, Ph.D Scholar

Mr Ch Ashutosh Swain, MCA

Mr Nrusingh P Achraya, MCA

Mr Devi P Kanungo, M.Tech Scholar

Mr M.K Sahu, M.Tech Scholar

Multi-objective Design Optimization of Three-Phase Induction

Motor Using NSGA-II Algorithm 1Soumya Ranjan and Sudhansu Kumar Mishra

A Comparative Study of Different Feature Extraction Techniques

for Offline Malayalam Character Recognition 9Anitha Mary M.O Chacko and P.M Dhanya

A Context Sensitive Thresholding Technique for Automatic

Image Segmentation 19Anshu Singla and Swarnajyoti Patra

Encryption for Massive Data Storage in Cloud 27Veeralakshmi Ponnuramu and Latha Tamilselvan

An Integrated Approach to Improve the Text Categorization

Using Semantic Measures 39

K Purna Chand and G Narsimha

An Android-Based Mobile Eye Gaze Point Estimation System

for Studying the Visual Perception in Children with Autism 49

J Amudha, Hitha Nandakumar, S Madhura, M Parinitha Reddy

and Nagabhairava Kavitha

FPGA Implementation of Various Image Processing Algorithms

Using Xilinx System Generator 59

M Balaji and S Allin Christe

A Study of Interestingness Measures for Knowledge Discovery

in Databases—A Genetic Approach 69Goyal Garima and Jyoti Vashishtha

xix

Efficient Recognition of Devanagari Handwritten Text 81Teja C Kundaikar and J.A Laxminarayana

Quality Assessment of Data Using Statistical and Machine

Learning Methods 89Prerna Singh and Bharti Suri

Design of Biorthogonal Wavelets Based on Parameterized Filter

for the Analysis of X-ray Images 99P.M.K Prasad, M.N.V.S.S Kumar and G Sasi Bhushana Rao

An Efficient Multi-view Based Activity Recognition System

for Video Surveillance Using Random Forest 111

J Arunnehru and M.K Geetha

Position and Orientation Control of a Mobile Robot

Using Neural Networks 123

D Narendra Kumar, Halini Samalla, Ch Jaganmohana Rao,

Y Swamy Naidu, K Alfoni Jose and B Manmadha Kumar

Fuzzy C-Means (FCM) Clustering Algorithm: A Decade Review

from 2000 to 2014 133Janmenjoy Nayak, Bighnaraj Naik and H.S Behera

Character Recognition Using Firefly Based Back Propagation

Neural Network 151M.K Sahoo, Janmenjoy Nayak, S Mohapatra, B.K Nayak

and H.S Behera

Analyzing Data Through Data Fusion Using Classification

Techniques 165Elizabeth Shanthi and D Sangeetha

Multi-objective Particle Swarm Optimization in Intrusion

Detection 175Nimmy Cleetus and K.A Dhanya

Vision Based Traffic Personnel Hand Gesture Recognition

Using Tree Based Classifiers 187

R Sathya and M Kalaiselvi Geetha

Optimization of the Investment Casting Process Using Genetic

Algorithm 201Sarojrani Pattnaik and Sutar Mihir Kumar

Cyclostationary Feature Detection Based Spectrum Sensing Technique

of Cognitive Radio in Nakagami-m Fading Environment 209Deborshi Ghosh and Srijibendu Bagchi

A Modified Real Time A* Algorithm and Its Performance Analysis

for Improved Path Planning of Mobile Robot 221P.K Das, H.S Behera, S.K Pradhan, H.K Tripathy and P.K Jena

Optimum Design and Performance Analysis of Dipole Planar

Array Antenna with Mutual Coupling Using Cuckoo Search

Algorithm 235Hrudananda Pradhan, Biswa Binayak Mangaraj and Iti Saha Misra

Model Based Test Case Generation from UML Sequence

and Interaction Overview Diagrams 247Ajay Kumar Jena, Santosh Kumar Swain and Durga Prasad Mohapatra

Enhancing HMM Based Malayalam Continuous Speech Recognizer

Using Artificial Neural Networks 259Anuj Mohamed and K.N Ramachandran Nair

Classification of Heart Disease Using Nạve Bayes and Genetic

Algorithm 269Santosh Kumar and G Sahoo

Solution for Traversal Vulnerability and an Encryption-Based

Security Solution for an Inter-cloud Environment 283

S Kirthica and Rajeswari Sridhar

Efficient Spread of Influence in Online Social Networks 293Gypsy Nandi and Anjan Das

Adaptive FIR Filter to Compensate for Speaker Non-linearity 301Varsha Varadarajan, Kinnera Pallavi, Gautam Balgovind

and J Selvakumar

A Pi-Sigma Higher Order Neural Network for Stock Index

Forecasting 311S.C Nayak, B.B Misra and H.S Behera

Comparison of Statistical Approaches for Tamil to English

Translation 321

R Rajkiran, S Prashanth, K Amarnath Keshav and Sridhar Rajeswari

An Integrated Clustering Framework Using Optimized K-means

with Firefly and Canopies 333

S Nayak, C Panda, Z Xalxo and H.S Behera

A Cooperative Intrusion Detection System for Sleep Deprivation

Attack Using Neuro-Fuzzy Classifier in Mobile Ad Hoc Networks 345Alka Chaudhary, V.N Tiwari and Anil Kumar

Improving the Performance of a Proxy Cache Using Expectation

Maximization with Naive Bayes Classifier 355

P Julian Benadit, F Sagayaraj Francis and U Muruganantham

Comparative Study of On-Demand and Table-Driven Routing

Protocols in MANET 369

G Kumar Pallai, S Meenakshi, A Kumar Rath and B Majhi

A Novel Fast FCM Clustering for Segmentation of Salt

and Pepper Noise Corrupted Images 381

B Srinivasa Rao and E Srinivasa Reddy

Non Linear Autoregressive Model for Detecting Chronic

Alcoholism 393Surendra Kumar, Subhojit Ghosh, Suhash Tetarway, Shashank Sawai,

Pillutla Soma Sunder and Rakesh Kumar Sinha

Theoretical Analysis of Expected Population Variance Evolution

for a Differential Evolution Variant 403

S Thangavelu, G Jeyakumar, Roshni M Balakrishnan

and C Shunmuga Velayutham

Quantification and 3D Visualization of Articular Cartilage

of Knee Joint Using Image Processing Techniques 417M.S Mallikarjunaswamy, Mallikarjun S Holi and Rajesh Raman

Application of Particle Swarm Optimization and User Clustering

in Web Search 427Sumathi Ganesan and Sendhilkumar Selvaraju

Analyzing Urban Area Land Coverage Using Image

Classification Algorithms 439

T Karthikeyan and P Manikandaprabhu

CALAM: Linguistic Structure to Annotate Handwritten Text Image

Corpus 449Prakash Choudhary and Neeta Nain

A Novel PSO Based Back Propagation Learning-MLP (PSO-BP-MLP)for Classification 461Himansu Das, Ajay Kumar Jena, Janmenjoy Nayak, Bighnaraj Naik

and H.S Behera

Quantum Based Learning with Binary Neural Network 473

Om Prakash Patel and Aruna Tiwari

Graphene Nano-Ribbon Based Schottky Barrier Diode

as an Electric Field Sensor 483Dipan Bandyopadhyay and Subir Kumar Sarkar

Dynamic Slicing of Object-Oriented Programs in Presence

of Inheritance 493S.R Mohanty, M Sahu, P.K Behera and D.P Mohapatra

Prediction of Heart Disease Using Classification Based Data

Mining Techniques 503Sujata Joshi and Mydhili K Nair

An Empirical Analysis of Software Reliability Prediction

Through Reliability Growth Model Using Computational

Intelligence 513Manmath Kumar Bhuyan, Durga Prasad Mohapatra, Srinivas Sethi

and Sumit Kar

A Harmony Search Based Gradient Descent Learning-FLANN

(HS-GDL-FLANN) for Classification 525Bighnaraj Naik, Janmenjoy Nayak, H.S Behera and Ajith Abraham

Improved AODV Performance in DOS and Black Hole

Attack Environment 541Anurag Gupta, Bhupendra Patel, Kamlesh Rana and Rahul Pradhan

Airfoil Self Noise Prediction Using Linear Regression Approach 551Shiju Sathyadevan and M.A Chaitra

Detection of Outliers in an Unsupervised Environment 563

M Ashwini Kumari, M.S Bhargavi and Sahana D Gowda

Discovering Latent Relationships Among Learning Style Dimensions

Using Association Rule Mining 575

C Beulah Christalin Latha, E Kirubakaran and Ranjit Jeba Thangaiah

Empirical Analysis of Low Power and High Performance

Multiplier 585

K Hemavathi and G Manmadha Rao

Facial Expression Recognition Using Modified Local Binary

Pattern 595Suparna Biswas and Jaya Sil

On Multigranular Approximate Rough Equivalence

of Sets and Approximate Reasoning 605B.K Tripathy, Prateek Saraf and S.Ch Parida

Spatial and Temporal Analysis of VOC Concentration in Capital City

of Odisha in Indian Sub-continent 617

S Rath, S.K Pandey, D.P Sandha, M Mohapatra, B Rath,

T Grahacharya and B.P.S Sahoo

Generating Prioritized Test Sequences Using Firefly Optimization

Technique 627Vikas Panthi and D.P Mohapatra

Directional Multi-scaled Fusion Based Median Filter for Removal

of RVIN 637Aparna Sarkar, Suvamoy Changder and J.K Mandal

Design of Area Optimized Sobel Edge Detection 647Sunil Kumar Kuppili and P.M.K Prasad

Accelerated FFT Computation for GNU Radio Using GPU

of Raspberry Pi 657

S Sabarinath, R Shyam, C Aneesh, R Gandhiraj and K.P Soman

Feature Extraction and Performance Analysis of EEG Signal

Using S-Transform 665Monorama Swain, Rutuparna Panda, Himansu Mahapatra

and Sneha Tibrewal

Image Super Resolution Reconstruction Using Iterative

Adaptive Regularization Method and Genetic Algorithm 675S.S Panda, G Jena and S.K Sahu

Semi-markov Process Based Cooperation Enforcement Mechanism

for MANETs 683

J Sengathir and R Manoharan

Reduction Combination Determination for Efficient Microarray

Data Classification with Three Stage Dimensionality

Reduction Approach 695Rasmita Dash and B.B Misra

Author Index 705

Editors ’ Biography

Prof Lakhmi C Jainis with the Faculty of Education, Science, Technology andMathematics at the University of Canberra, Australia and University of SouthAustralia, Australia He is a Fellow of the Institution of Engineers, Australia.Professor Jain founded the Knowledge-Based Intelligent Engineering System(KES) International, a professional community for providing opportunities forpublication, knowledge exchange, cooperation, and teaming Involving around5,000 researchers drawn from universities and companies worldwide, KES facili-tates international cooperation and generates synergy in teaching and research KESregularly provides networking opportunities for the professional communitythrough one of the largest conferences of its kind in the area of KES His interestsfocus on artificial intelligence paradigms and their applications in complex systems,security, e-education, e-healthcare, unmanned air vehicles, and intelligent agents.Prof Himansu Sekhar Behera is working as a Reader in the Department ofComputer Science Engineering and Information Technology, Veer Surendra SaiUniversity of Technology (VSSUT) (A Unitary Technical University, Established

by Government of Odisha), Burla, Odisha He has received M.Tech in ComputerScience and Engineering from N.I.T, Rourkela (formerly R.E.C., Rourkela) andDoctor of Philosophy in Engineering (Ph.D.) from Biju Pattnaik University ofTechnology (BPUT), Rourkela, Government of Odisha respectively He has pub-lished more than 80 research papers in various international journals and confer-ences, edited 11 books and is acting as a member of the editorial/reviewer board ofvarious international journals He is proficient in the field of Computer ScienceEngineering and served in the capacity of program chair, tutorial chair, and acted asadvisory member of committees of many national and international conferences.His research interest includes Data Mining and Intelligent Computing He isassociated with various educational and research societies like OITS, ISTE, IE,ISTD, CSI, OMS, AIAER, SMIAENG, SMCSTA, etc He is currently guidingseven Ph.D scholars

xxvii

Prof Jyotsna Kumar Mandalis working as Professor in Computer Science andEngineering, University of Kalyani, India Ex-Dean Faculty of Engineering,Technology and Management (two consecutive terms since 2008) He has 26 years

of teaching and research experiences He was Life Member of Computer Society ofIndia since 1992 and life member of Cryptology Research Society of India, member

of AIRCC, associate member of IEEE and ACM His research interests includeNetwork Security, Steganography, Remote Sensing and GIS Application, ImageProcessing, Wireless and Sensor Networks Domain Expert of Uttar Banga KrishiViswavidyalaya, Bidhan Chandra Krishi Viswavidyalaya for planning and inte-gration of Public domain networks He has been associated with national andinternational journals and conferences The total number of publications to hiscredit is more than 320, including 110 publications in various international journals.Currently, he is working as Director, IQAC, Kalyani University

Prof Durga Prasad Mohapatra received his Ph.D from Indian Institute ofTechnology Kharagpur and is presently serving as an Associate Professor in NITRourkela, Odisha His research interests include software engineering, real-timesystems, discrete mathematics, and distributed computing He has published morethan 30 research papers in these fields in various international Journals and con-ferences He has received several project grants from DST and UGC, Government

of India He received the Young Scientist Award for the year 2006 from OrissaBigyan Academy He has also received the Prof K Arumugam National Awardand the Maharashtra State National Award for outstanding research work in Soft-ware Engineering for the years 2009 and 2010, respectively, from the IndianSociety for Technical Education (ISTE), New Delhi He is nominated to receive theBharat Shiksha Ratan Award for significant contribution in academics awarded bythe Global Society for Health and Educational Growth, Delhi

of Three-Phase Induction Motor Using

NSGA-II Algorithm

Soumya Ranjan and Sudhansu Kumar Mishra

Abstract The modeling of electrical machine is approached as a system zation, more than a simple machine sizing Hence wide variety of designs areavailable and the task of comparing the different options can be very difficult

optimi-A number of parameters are involved in the design optimization of the inductionmotor and the performance relationship between the parameters also is implicit Inthis paper, a multi-objective problem is considered in which three phase squirrelcage induction motor (SCIM) has been designed subject to the efficiency and powerdensity as objectives The former is maximized where the latter is minimizedsimultaneously considering various constraints Three single objective methodssuch as Tabu Search (TS), Simulated Annealing (SA) and Genetic Algorithm (GA)

is used for comparing the Pareto solutions Performance comparison of techniques

is done by performing different numerical experiments The result shows thatNSGA-II outperforms other three for the considered test cases

Keywords Multi-objective optimization
Induction motors
Multi-objectiveevolutionary algorithms
Single objective evolutionary algorithm

1 Introduction

Three-phase induction motors have been widely used in industrial applications.Over the past decade, there have been clear areas in motor utilization that demandhigher power density and increased energy efficiency In many industrial

S Ranjan ( &)

Department of Electrical and Electronics Engineering, NIST, Berhampur, India

e-mail: Soumya.biteee@gmail.com

S.K Mishra

Department of Electrical and Electronics Engineering, Birla Institute of Technology,

Mesra, Ranchi, India

e-mail: Sudhansu.nit@gmail.com

© Springer India 2015

L.C Jain et al (eds.), Computational Intelligence in Data Mining - Volume 2,

Smart Innovation, Systems and Technologies 32, DOI 10.1007/978-81-322-2208-8_1

1

applications, motor size and inertia are critical Motors with high power density canoffer a performance advantage in applications such as paper machines However,high-power density cannot compromise reliability and efficiency In such multi-objective optimization (MO), it is impossible to obtain the solution with maxi-mizing or minimizing all objectives simultaneously because of the trade off relationbetween the objectives When the MO is applied to the practical design process, it isdifficult to achieve an effective and robust optimal solution within an acceptablecomputation time The solutions obtained are known as Pareto-optimal solutions ornon-dominated solutions The rest is called dominated solutions There are severalmethods to solve MO problems and one method of them, Pareto optimal solutionsare generally used for the balanced solutions between objectives.

Appelbaum proposed the method of“boundary search along active constrains”

in 1987 [1] Madescu proposed the nonlinear analytical iterativefield-circuit model(AIM) in 1996 by Madescu et al [2] However, these techniques have manyshortcomings to provide fast and accurate solution, particularly when the optimalsolution to a problem has many variables and constraints Thus, to deal with suchdifficulties efficient optimization strategies are required This can be overcome bymulti-objective optimization (MO) technique [3–7]

This paper aims at MO which incorporates NSGA-II algorithm for minimization

of power density and maximization of efficiency of three phases SCIM using ferent nonlinear constrained optimization techniques [8–10] The Pareto-optimiza-tion technique is used in order to solve the multi-objective optimization problem ofelectric motor drive in a parametric fashion It results in a set of optimal solutionsfrom which an appropriate compromise design can be chosen based on the pref-erence of the designer In addition to that various SOEA techniques such as Sim-ulated Annealing (SA), Tabu Search (TS), Genetic Algorithm (GA) is applied tocompare among Pareto-optimal solutions [11] Their performance has been eval-uated by the metrics such as Delta, Convergence (C) and Spacing (S) throughsimulation studies

dif-2 Multi-objective Optimization Design

The general formulation of MOPs as [12]

Maximize/Minimize

fð~xÞ ¼ fð1ð~xÞ; f2ð~xÞ; ; fMð~xÞÞ ð1ÞSubjected to constraints:

where~x represents a vector of decision variables ~x ¼ xf ; x ; ; x gT:

The search space is limited by

is said todominate over~v ¼ fv1; v2; ; vkgT

3 Multi-objective Evolutionary Algorithm Frameworks

A majority of MOEAs in both the research and the application areas are dominance based which are mostly the same frameworks as that of NSGA-II Inthese algorithms a selection operator based on Pareto-domination and a reproduc-tion operator is used The operator of the MOEAs guides the population iterativelytowards non-dominated regions by preserving the diversity to get the Pareto-optimalset The evaluate operator leads to population convergence towards the efficientfrontier and helps preserve the diversity of solutions along the efficient frontier.Both goals are achieved by assigning a rank and a density value to each solution.The MOEAs providefirst priority to non-dominance and second priority to diver-sity However, the methods by which they achieve these two fundamental goalsdiffer The main difference between the algorithms lies in theirfitness assignmenttechniques Coello et al Classifies the constraints handling methods into five cat-egories: (1) penalty functions (2) special representations and operators (3) repairalgorithms (4) separate objective and constraints and (5) hybrid methods [15,16]

Pareto-4 Design Optimization of Induction Motor

In this paper the design of induction motor is formulated by MOEAs based on dominated sorting, NSGA-II which does not combine the two objectives to obtainthe Pareto-optimal solution set Here, the two objectives are taken individually and

non-an attempt is made to optimize both simultnon-aneously The main objective is tomaximize efficiency (η) and minimize power density (ξ) The proposed NSGA-II issuitably oriented in such a way as to optimize the two objectives To express boththe objectives in maximization form, the first objective ξ is expressed as −ξ Inaddition to these objectives, different practical constraints mentioned are alsoconsidered In order to design, the problem is expressed as Maximize η and –ξsimultaneously considering all constraints [17,18].

The sizing equation of an induction machine is

PRð Þ ¼IM

ffiffiffi2

p

p22ð1 þ K/ÞKxg cos /rBgAf

pk2

In terms of efficiency (η) can be written as

g ¼ PRðIMÞ2ð1 þ K/Þffiffiffi

f IMð Þ ¼

ffiffiffi2

p

p22ð1 þ K/ÞKxg cos /rBgAf

pk2 0

5 Performance Measure for Comparison

The final Pareto-optimal front obtained from different MOEAs techniques iscompared using performance metrics such as Spacing (S), Diversity metric (Δ),Convergence metric (C) [17] These performance metrics set the benchmark tocompare the results and select the best outcomes

6 Simulation Results

The 5 kW, 4-pole, three-phase squirrel-cage induction motor is chosen as a sampledesign The rated frequency is 50 Hz and voltage is 170 V Also, the ratio ofmaximum torque to nominal torque is set 2.5 as a constraint Lower limit of

efficiency is 90 % and that of power density is 0.3 kW/kg The population size is set

to be 100 The algorithms stop after 20,000 function evaluations Initial populationsare generated by uniformly randomly sampling from the feasible search space Theuniform Crossover rate is taken 0.8 The mutation rate is 0.10 where it is taken as1/n, i.e n is 10, the number of decision variables

Table1 shows the S metric and Δ metric obtained using all four algorithms.Table1 shows that the S andΔ metric value for NSGA-II is less than other threealgorithms and hence its performance is better among all

Table 2 shows the result obtained for Convergence (C) metrics The values0.5988 in the fourth row,first column means almost all solutions from final pop-ulations obtained by NSGA-II dominates the solutions obtained by SA The values

0 in thefirst row, first column mean that no solution of the non-dominated lation obtained by TS, GA and NSGA-II is dominated by solutions from finalpopulations obtained by SA From the result, it clear that the performance ofNSGA-II significantly outperforms the competing algorithms in the consideredoptimal design of induction motor

popu-The comparison time computed by the CPU is shown in Table3 The mean timeand the variance (var) of time for NSGA-II algorithm is less than other algorithms.The Simulation statistics generated by the four algorithms NSGA-II, GA, TS, SArespectively are illustrated from Figs.1,2,3and4 It is shown in Fig.5that NSGA-

II results in wide areas of convergence and is diversified

Table 1 The performance

Std 2.08E −1 1.93E −1 1.48E −1 1.47E −1

Table 2 The result obtained

Table 3 Comparison of CPU

CPU time Mean Var Mean Var Mean Var Mean Var

Fig 1 Plots of Pareto fronts achieved by NSGA II

Fig 2 Plots of Pareto fronts achieved by GA

Fig 3 Plots of Pareto fronts achieved by TS

7 Conclusion

In this paper, the multi-objective design optimization based on NSGA-II and sizeequations are applied for the three phase induction motors In order to effectivelyobtain a set of Pareto optimal solutions, ranking method is applied From theresults, we can select the balanced optimal solution between the power density and

efficiency In case of optimized model, the efficiency increases at 80 % and thepower density is also increased 12 kW/kg, compared to the SA, TS and GA result ofthe initial model The performance metrics of NSGA-II results in best possiblePareto solutions The proposed method can be efficiently and effectively used tomulti-objectives design optimization of the machine cost and efficiency of electricmachines

Fig 4 Plots of Pareto fronts achieved by SA

Fig 5 Pareto front at different cardinality

3 Appelbaum, J., Khan, I.A., Fuchs, E.F.: Optimization of three-phase induction motor design with respect to ef ficiency In: Proceedings of ICEM, pp 639–642, Lousanne, Switzerland (1984)

4 Ramarathnam, R., Desai, B.G.: Optimization of poly-phase induction motor design —a nonlinear programming approach IEEE Trans Power Apparatus Syst PAS-90, 570 –578 (1971)

5 Fetih, N.H., El-Shewy, H.M.: Induction motor optimum design, including active power loss effect IEEE Trans Energy Convers 1(3), 155 –160 (1986)

6 Deb, K.: Multi-Objective Optimization Using Evolutionary Algorithms, pp 48 –55 Wiley, New York (2001)

7 Fuchs, E.F., Appelbaum, J., Khan, I.A, Holl, J., Frank, U V.: Optimization of induction motor

ef ficiency, vol 1 Three Phase Induction Motors, Final Report, EPRIEL-4152-ccm, Project, Colorado (1985)

8 Deb, K., Pratap, A., Agarwal, S., Meyarivan, T.: A fast and elitist multiobjective genetic algorithm: NSGA-II IEEE Trans Evol Comput 6, 182 –197 (2002)

9 Deb, K., Anand, A., Joshi, D.: A computationally ef ficient evolutionary algorithm for parameter optimization Evol Comput 10(4), 371 –395 (2002)

real-10 Ranjan, S., Mishra, S.K., Behera, S.K.: A comparative performance evaluation of evolutionary algorithms for optimal design of three-phase induction motor In: IEEE Xplore, pp 1 –5 (2014)

11 Ranjan, S., et al.: Multiobjective optimal design of three-phase induction motor for traction systems using an asymmetrical cascaded multilevel inverter IJISM 19 –125 (2014)

12 Coello, C.A.C., Gregorio, T.P., Maximino, S.L.: Handling multiple objectives with particle swarm optimization IEEE Trans Evol Comput 8(3), 256 –279 (2004)

13 Pareto, V., Cours, D.: Economie Politique, vol 8 (1986)

14 Deb, K., Pratap, A., Agarwal, S., Meyarivan, T.: A fast and elitist multiobjective genetic algorithm: NSGA-II IEEE Trans Evol Comput 6(2), 182 –197 (2002)

15 Zhou, A., Qu, B.-Y., Li, H., Zhao, S.-Z., Suganthan, P.N., Zhang, Q.: Multiobjective evolutionary algorithms: a survey of the state-of-the-art Swarm Evol Comput 1(1), 32 –49 (2011)

16 Montes, E.M., Coello, C.A.C.: Constraint-handling in nature-inspired numerical optimization: past, present and future Swarm Evol Comput 1(4), 173 –194 (2011)

17 Huang, S., Luo, J., Leonardi, F., Lipo, T.A.: A general approach to sizing and power density equations for comparison of electrical machines IEEE Trans Ind Appl 34(1), 92 –97 (1998)

18 Kim, M.K., Lee, C.G., Jung, H.K.: Multiobjective optimal design of three-phase induction motor using improved evolution strategy IEEE Trans Magn 34(5), 2980 –2983 (1998)

19 Ojo, O.: Multiobjective optimum design of electrical machines for variable speed motor drives Tennesse Technological University, Cookeville, TN 38505

Extraction Techniques for Of fline

Malayalam Character Recognition

Anitha Mary M.O Chacko and P.M Dhanya

Abstract Offline Handwritten Character Recognition of Malayalam scripts havegained remarkable attention in the past few years The complicated writing style ofMalayalam characters with loops and curves make the recognition process highlychallenging This paper presents a comparative study of Malayalam characterrecognition using 4 different feature sets—Zonal features, Projection histograms,Chain code histograms and Histogram of Oriented Gradients The performance ofthese features for isolated Malayalam vowels and 5 consonants are evaluated in thisstudy using feedforward neural networks as classifier The final recognition resultswere computed using a 5 fold cross validation scheme The best recognitionaccuracy of 94.23 % was obtained in this study using Histogram of OrientedGradients features

Keywords Offline character recognition
Feature extraction
Neural networks

1 Introduction

Offline character recognition is the process of translating handwritten text fromscanned, digitized or photographed images into a machine editable format Com-pared to online recognition, offline recognition is a much more challenging task duethe lack of temporal and spatial information Character recognition research hasgained immense popularity because of its potential applications in the areas ofpostal automation, bank check processing, number plate recognition etc Eventhough ambient studies have been performed in foreign languages [1], only very

A.M.M.O Chacko ( &)
P.M Dhanya

Department of Computer Science and Engineering, Rajagiri School of Engineering

and Technology, Kochi, India

e-mail: anithamarychacko@gmail.com

P.M Dhanya

e-mail: dhanya_pm@rajagiritech.ac.in

© Springer India 2015

L.C Jain et al (eds.), Computational Intelligence in Data Mining - Volume 2,

Smart Innovation, Systems and Technologies 32, DOI 10.1007/978-81-322-2208-8_2

9

few works exist in the Malayalam character recognition domain This is mainly due

to its extremely large character set and complicated writing style with loops curvesand holes

Some of the major works reported in the Malayalam character recognitiondomain are as follows: Lajish [2] proposed thefirst work in Malayalam OCR usingfuzzy zoning and normalized vector distances 1D wavelet transform of vertical andhorizontal projection profiles were used in [3] for the recognition of Malayalamcharacters The performance of wavelet transform of projection profiles using 12different wavelet filters were analyzed in [4] In [5], recognition of Malayalamvowels was done using chain code histogram and image centroid They have alsoproposed another method for Malayalam character recognition using Haar wavelettransform and SVM classifier [6] Moni and Raju used Modified Quadratic Clas-

sifier and 12 directional gradient features for handwritten Malayalam characterrecognition [7] Here gradient directions were computed using Sobel operators andwere mapped into 12 directional codes Recently, Jomy John proposed anotherapproach for offline Malayalam recognition using gradient and curvature calcula-tion and dimensionality reduction using Principal Component Analysis (PCA) [8]

A detailed survey on Malayalam character recognition is presented in [9]

A general handwritten character recognition system consists of mainly 4phases—Preprocessing, Feature Extraction, Classification and Postprocessing.Among these, feature extraction is an important phase that determines the recog-nition performance of the system To get an idea of recognition results of differentfeature extraction techniques in Malayalam character recognition, we have per-formed a comparative study using 4 different features—Zonal features, projectionhistograms, chain codes and Histogram of Oriented Gradients (HOG) features Theperformance of these four feature sets are analyzed by using a two layer feedfor-ward neural network as classifier

The paper is structured as follows: Sect.2 presents the data collection methodused and the sequence of preprocessing steps done Section3describes the featureextraction procedure for the four feature sets The classifier used is introduced inSect 4 Section 5 presents the experimental results and discussions and finallyconclusion is presented in Sect.6

2 Data Collection and Preprocessing

Malayalam belongs to the Dravidian family of languages which has official guage status in Kerala The complete character set of Malayalam consists of 15vowels, 36 consonants, 5 chillu, 9 vowel signs, 3 consonant signs, 3 specialcharacters and 57 conjunct consonants Since a benchmarking database is notavailable for Malayalam, we have created a database of 260 samples for the isolatedMalayalam vowels and 5 consonants (‘ka’,‘cha’,‘tta’,‘tha’ and ‘pa’) For this 13class recognition problem, we have collected handwritten samples from 20 people

lan-belonging to different age groups and professions Each of these 13 characters areassigned class-ids The scanned images were then subjected to preprocessing.Figure1 shows sample characters of the database.

2.1 Preprocessing

Preprocessing steps are carried out to reduce variations in the writing style ofdifferent people The sequences of preprocessing steps (Fig 2) carried out are asfollows: Here, scanned images are binarized using Otsu’s method of global thres-holding This method is based on finding the threshold that minimizes theintra-class variance A large amount of noise such as salt and pepper noise mayexist in the image acquired by scanning So in order to reduce this noise to someextent, we have applied a 3 × 3 median filter In the segmentation process, theFig 1 Samples of handwritten Malayalam characters

Fig 2 Preprocessing steps.

a Scanned image b Binarized

image c Size normalized

image

character images are separated into individual text lines from which characters areisolated using connected component labeling Finally, the images are resized to

256 × 256 using bicubic interpolation techniques This operation ensures that allcharacters have a predefined height and width

3 Feature Extraction

The performance of an HCR system depends to a great extent on the extractedfeatures Over the years, many feature extraction techniques have been proposed forcharacter recognition A survey of feature extraction techniques is presented in [10]

In this study, we have used 4 sets of features for comparing the performance of thecharacter recognition system: Zonal features, Projection histograms, Chain codehistograms and Histogram of Oriented Gradients

3.1 Zoning

Zoning is a popular method used in character recognition tasks In this method, thecharacter images are divided into zones of predefined sizes and then features arecomputed for each of these zones Zoning obtains local characteristics of an image.Here, we have divided the preprocessed character images into 16 zones (4× 4) as inand then pixel density features were computed for each of the zones (Fig.3) Theaverage pixel density was calculated by dividing the number of foreground pixels

by the total number of pixels in each zone i

dðiÞ ¼Number of foreground pixels in zone i

Fig 3 4 × 4 zoning

Thus we have obtained 16 density features which are used as input to theclassifier.

3.2 Projection Pro file

Projection profile is an accumulation of black pixels along rows or columns of animage The discriminating power of horizontal and vertical projection profiles makethem well suitable for the recognition of a complex language like Malayalam.Projection profiles have been successfully applied for Malayalam character rec-ognition [3,4]

In this study, we have extracted both vertical and horizontal projection profiles

by counting the pixels column wise and row wise respectively which together forms

a 512 dimension feature vector (Fig.4shows the vertical and horizontal projectionhistogram for a Malayalam character‘tha’)

Since, the size of the feature vector is too large, we have applied PrincipalComponent Analysis (PCA) to reduce the dimensionality of the feature set PCA is

a technique that reduces the dimensionality of the data while retaining as muchvariations as possible in the original dataset Using PCA, we have reduced thedimension of the feature vector from 512 to 260

3.3 Chain Code Features

The chain code approach proposed by Freeman [11] is a compact way to representthe contour of an object The chain codes are computed by moving along theboundary of the character in clockwise/anticlockwise direction and assigning each

Fig 4 Projection histogram of character ‘tha’ a Horizontal projection histogram b Vertical projection histogram