Advances in Intelligent Systems and Computing 343

Durbadal MandalRajib KarSwagatam DasBijaya Ketan Panigrahi Editors

Intelligent Computing and ApplicationsProceedings of the International Conference on ICA, 22–24 December 2014

Advances in Intelligent Systems and Computing

Volume 343

Series editor

Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Polande-mail: kacprzyk@ibspan.waw.pl

About this Series

The series “Advances in Intelligent Systems and Computing” contains publications on theory,applications, and design methods of Intelligent Systems and Intelligent Computing. Virtuallyall disciplines such as engineering, natural sciences, computer and information science, ICT,economics, business, e-commerce, environment, healthcare, life science are covered. The listof topics spans all the areas of modern intelligent systems and computing.

The publications within “Advances in Intelligent Systems and Computing” are primarilytextbooks and proceedings of important conferences, symposia and congresses. They coversignificant recent developments in the field, both of a foundational and applicable character.An important characteristic feature of the series is the short publication time and world-widedistribution. This permits a rapid and broad dissemination of research results.

Advisory Board

Chairman

Nikhil R. Pal, Indian Statistical Institute, Kolkata, Indiae-mail: nikhil@isical.ac.in

Members

Rafael Bello, Universidad Central “Marta Abreu” de Las Villas, Santa Clara, Cubae-mail: rbellop@uclv.edu.cu

Emilio S. Corchado, University of Salamanca, Salamanca, Spaine-mail: escorchado@usal.es

Hani Hagras, University of Essex, Colchester, UKe-mail: hani@essex.ac.uk

László T. Kóczy, Széchenyi István University, Győr, Hungarye-mail: koczy@sze.hu

Vladik Kreinovich, University of Texas at El Paso, El Paso, USAe-mail: vladik@utep.edu

Chin-Teng Lin, National Chiao Tung University, Hsinchu, Taiwane-mail: ctlin@mail.nctu.edu.tw

Jie Lu, University of Technology, Sydney, Australiae-mail: Jie.Lu@uts.edu.au

Patricia Melin, Tijuana Institute of Technology, Tijuana, Mexicoe-mail: epmelin@hafsamx.org

Nadia Nedjah, State University of Rio de Janeiro, Rio de Janeiro, Brazile-mail: nadia@eng.uerj.br

Ngoc Thanh Nguyen, Wroclaw University of Technology, Wroclaw, Polande-mail: Ngoc-Thanh.Nguyen@pwr.edu.pl

Jun Wang, The Chinese University of Hong Kong, Shatin, Hong Konge-mail: jwang@mae.cuhk.edu.hk

More information about this series at http://www.springer.com/series/11156

Durbadal Mandal • Rajib KarSwagatam Das • Bijaya Ketan PanigrahiEditors

Intelligent Computingand ApplicationsProceedings of the International Conferenceon ICA, 22–24 December 2014

123

EditorsDurbadal MandalDepartment of Electronics andCommunication Engineering

NITDurgapur, West BengalIndia

Rajib KarDepartment of Electronics andCommunication Engineering

NITDurgapur, West BengalIndia

Swagatam DasMachine Intelligence UnitISIKolkata, West BengalIndia

Bijaya Ketan PanigrahiDepartment of Electrical EngineeringIIT DelhiNew DelhiIndia

ISSN 2194-5357 ISSN 2194-5365 (electronic)Advances in Intelligent Systems and ComputingISBN 978-81-322-2267-5 ISBN 978-81-322-2268-2 (eBook)DOI 10.1007/978-81-322-2268-2

Library of Congress Control Number: 2015930027

Springer New Delhi Heidelberg New York Dordrecht London© Springer India 2015This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or partof 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 transmissionor information storage and retrieval, electronic adaptation, computer software, or by similar ordissimilar methodology now known or hereafter developed.The use of general descriptive names, registered names, trademarks, service marks, etc. in thispublication does not imply, even in the absence of a specific statement, that such names are exemptfrom 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 thisbook are believed to be true and accurate at the date of publication. Neither the publisher nor theauthors or the editors give a warranty, express or implied, with respect to the material containedherein 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)

Preface

This AISC volume contains the papers presented at the First InternationalConference on Intelligent Computing and Application (ICICA 2014) held duringDecember 22–24, 2014, at NIT Durgapur, West Bengal, India. ICICA 2014 is thefirst international conference aiming at bringing together researchers from academiaand industry to report and review the latest progresses in the cutting-edge researchon various research areas of Image, Language Processing, Computer Vision andPattern Recognition, Machine Learning, Data Mining and Computational LifeSciences, Management of Data including Big Data and Analytics, Distributed andMobile Systems including Grid and Cloud infrastructure, Information Security andPrivacy, VLSI, Electronic Circuits, Power Systems, Antenna, Computational fluiddynamics and Heat transfer, Intelligent Manufacturing, Signal Processing, Intelli-gent Computing, Soft Computing, Bio-informatics, Bio Computing, Web Security,Privacy and E-Commerce, E-governance, Service Orient Architecture, Data Engi-neering, Open Systems, Optimization, Communications, Smart wireless and sensorNetworks, Smart Antennae, Networking and Information security, MachineLearning, Mobile Computing and Applications, Industrial Automation and MES,Cloud Computing, Green IT, IT for Rural Engineering, Business Computing,Business Intelligence, ICT for Education for solving hard problems, and finally tocreate awareness about these domains to a wider audience of practitioners.

ICICA 2014 received 144 paper submissions including two foreign countriesacross the globe. After a rigorous peer-review process involving 600 reviews, 70full length articles were accepted for oral presentation at the conference. Thiscorresponds to an acceptance rate of 48.61 % and is intended to maintain the highstandards of the conference proceedings. The papers included in this AISC volumecover a wide range of topics in intelligent computing and algorithms and their real-world applications in problems from diverse domains of science and engineering.

The conference featured distinguished keynote speakers: Dr. P.N. Suganthan,NTU, Singapore; Dr. Swagatam Das, ISI, Kolkata, India; Prof. B.K. Panigrahi, IITDelhi, India and Prof. B. Sikdar, Rensselaer Polytechnique Institute, USA. We takethis opportunity to thank the authors of the submitted papers for their hard work,adherence to the deadlines, and patience with the review process.

v

The quality of a referred volume depends mainly on the expertise and dedicationof the reviewers. We are indebted to the Program Committee/Technical Committeemembers, who produced excellent reviews in short time frames. We would also liketo thank TEQIP II for providing all the logistical support and financial assistance.First, we are indebted to NIT Durgapur, India for supporting our cause andencouraging us to organize the conference there. In particular, we would like toexpress our heartfelt thanks for providing us with the necessary financial supportand infrastructural assistance to hold the conference. Our sincere thanks to Prof.P. Gupta, Dean Research and Consultancy; Prof. D.K. Mandal; Prof. J.P. Sarkar;Prof. B. Haldar; Col. (Retd.) P.S. Sandhu (Registrar); Prof. A.K. Bhattacharjee;Prof. R. Ghatak; Prof. K.C. Ghanta; Prof. N.K. Roy; Dr. (Mrs.) S. Roy; Prof.S. Ghosh; Mr. A.K. Chattopadhyay; Mr. U.C. Mukherjee; Mr. A. Kumar; Mr. A.K.Bhagat and Prof. T. Kumar (Director, NIT Durgapur), for their encouragement andcontinuous support. We thank Maurice Clerc, France; Roderich Gross, England;Kalyanmoy Deb, India; Saman Halgamuge, Australia; Jeng-Shyang Pan, Taiwan;Peng Shi, UK; Javier Del Ser, Spain; Leandro Dos Santos Coelho, Brazil; S.S.Pattanaik, India; Gerardo Beni, USA; K. Parsopoulos, Greece; Lingfeng Wang,China; Athanasios V. Vasilakos, Athens; Pei-Chann Chang, Taiwan; Chilukuri K.Mohan, USA; Saeid Nahavandi, Australia; Abbas Khosravi, Australia; AlmoatazYoussef Abdelaziz, Egypt; K.T. Chaturvedi, India; M.K. Tiwari, India; Yuhui Shi,China; A. Anand, India; Dipankar Dasgupta, USA; Lakhmi Jain, Australia; X.Z.Gao, Finland; Juan Luis Fernandez Martinez, Spain; Oscar Castillo, Mexico; HeitorSilverio Lopes, Brazil; S.K. Udgata, India; Namrata Khemka, USA; G.K. Vena-yagamoorty, USA; Zong Woo Geem, USA; Ying Tan, China; S.G. Ponnambalam,Malaysia and Halina Kwasnicka, Poland for providing valuable guidelines andinspiration to overcome various difficulties in the process of organizing thisconference.

We would also like to thank the participants of this conference. Finally, we thankall the volunteers who made great efforts in meeting the deadlines and arrangingevery detail to make sure that the conference could run smoothly. We hope thereaders of these proceedings find the papers inspiring and enjoyable.

December 2014 Durbadal MandalRajib Kar

Swagatam DasBijaya Ketan Panigrahi

P.N. Suganthan

vi Preface

Conference Organizing Committee

Chief Patron

Prof. Bikas Sinha, Chairman, NIT Durgapur, India

Patron

Prof. T. Kumar, Director, NIT Durgapur, India

General Chair

Dr. D. Mandal, Dr. R. Kar, NIT Durgapur, India

General Co-chair

Prof. S. Das, ISI Kolkata, IndiaProf. B.K. Panigrahi, IIT Delhi, India

vii

Program Chair

Prof. D.K. Pratihar, IIT Kharagpur, IndiaProf. Jaya Sil, IIEST Shibpur, IndiaProf. S. Kundu, NIT Durgapur, India

Finance Chair

Dr. Rajat Mahapatra, NIT Durgapur, India

Publication Chairs

Prof. S.P. Ghoshal, Dr. A.K. Mal, NIT Durgapur, India

International Advisory Committee/Technical Committee

Maurice Clerc, FranceRoderich Gross, EnglandKalyanmoy Deb, IndiaSaman Halgamuge, AustraliaJeng-Shyang Pan, TaiwanPeng Shi, UKJavier Del Ser, SpainLeandro Dos Santos Coelho, BrazilS.S. Pattanaik, IndiaGerardo Beni, USAK. Parsopoulos, GreeceLingfeng Wang, ChinaAthanasios V. Vasilakos, AthensPei-Chann Chang, TaiwanChilukuri K. Mohan, USASaeid Nahavandi, AustraliaAbbas Khosravi, AustraliaAlmoataz Youssef Abdelaziz, EgyptK.T. Chaturvedi, IndiaM.K. Tiwari, IndiaYuhui Shi, ChinaA. Anand, India

viii Conference Organizing Committee

Dipankar Dasgupta, USALakhmi Jain, AustraliaX.Z. Gao, FinlandJuan Luis Fernandez Martinez, SpainOscar Castillo, MexicoHeitor Silverio Lopes, BrazilS.K. Udgata, IndiaNamrata Khemka, USAG.K. Venayagamoorty, USAZong Woo Geem, USAYing Tan, ChinaS.G. Ponnambalam, MalaysiaHalina Kwasnicka, Poland

Organizing Committee

Prof. P. Gupta, NIT DurgapurProf. D.K. Mandal, NIT DurgapurProf. J.P. Sarkar, NIT DurgapurProf. B. Haldar, NIT DurgapurCol. (Retd.) P.S. Sandhu, NIT DurgapurProf. A.K. Bhattacharjee, NIT DurgapurProf. R. Ghatak, NIT DurgapurProf. K.C. Ghanta, NIT DurgapurProf. N.K. Roy, NIT DurgapurDr. (Mrs.) S. Roy, NIT DurgapurProf. S. Ghosh, NIT DurgapurMr. A.K. Chattopadhyay, NIT DurgapurMr. U.C. Mukherjee, NIT DurgapurMr. A. Kumar, NIT DurgapurMr. A.K. Bhagat, NIT Durgapur

Conference Organizing Committee ix

Contents

Characterization of Fuzzy Tree Searches: A Perspective Note . . . . . . . 1Parag Bhalchandra, Santosh Khamitkar, Nilesh Deshmukh,Sakharam Lokhande and Satish Mekewad

A Modified SOM-Based RBFN for Rotation Invariant Clearand Occluded Fingerprint Recognition . . . . . . . . . . . . . . . . . . . . . . . . 11Sumana Kundu and Goutam Sarker

An Unsupervised OCA-based RBFN for Clear and OccludedFace Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19Dhananjoy Bhakta and Goutam Sarker

Analysing the Impact of Human Behaviour over MachineLearning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27Shailendra Singh Yadav and Bhupendra Verma

An Improved RC4 with Statistical Analysis on Ciphertexts . . . . . . . . . 37Suman Das, Hemanta Dey and Ranjan Ghosh

Gathering of Swarm of Mobile Robots in Presenceof Horizontal Obstacles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49U. Baveenther and Deepanwita Das

Design of Wireless Sensor Node to Measure Vibrationand Environment Parameter for Structural HealthMonitoring Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59Sen Siddheswar, Swarnakar Biplab and Datta Uma

Economic Load Dispatch Considering Non-smooth CostFunctions Using Predator–Prey Optimization . . . . . . . . . . . . . . . . . . . 67Sunanda Hazra and Provas Kumar Roy

xi

Sparse Denoising in Speckle Noise: A Comparative Studyof Dictionaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79Suchismita Maiti, Amish Kumar and Debashis Nandi

Compact RS(32, 28) Encoder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89Jagannath Samanta, Jaydeb Bhaumik and Soma Barman

Graphene–Silicene Composite can Increase the Efficiencyof Cloud Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97Dhiman Kumar Das and Sukadev Sahoo

Object-oriented Modeling of IDEA for E-learning Security . . . . . . . . . 105Ambalika Ghosh and Sunil Karforma

An Improved BAT-Optimized Cluster-Based Routingfor Wireless Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115Koteswararao Seelam, M. Sailaja and T. Madhu

An Hybrid Ant Routing Algorithm for Reliable ThroughputUsing MANET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127N. Umapathi, N. Ramaraj, D. Balasubramaniam and R. Adlin mano

Development of Type 2 Fuzzy Rough Ontology-basedMiddleware for Context Processing in Ambient SmartEnvironment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137A.B. Karthick Anand Babu and R. Sivakumar

Multiple-Image Encryption Using Genetic Algorithm . . . . . . . . . . . . . 145Subhajit Das, SatyendraNath Mandal and Nabin Ghoshal

Knowledge-Based System Architecture on CBR for Detectionof Cholera Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155Souvik Chakraborty, Chiranjit Pal, Shambo Chatterjee,Baisakhi Chakraborty and Nabin Ghoshal

Efficient Adaptive Routing Algorithm for WSNswith Sink Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167Ashwini V. Mannapur, Jayashree D. Mallapur and S.P. Parande

Detection of Degree of Sickness of Affected Eye Using FuzzyLogic and Histogram Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177Manisha Barman, J. Paul Choudhury and S. Biswas

xii Contents

Element Spacing Optimization of Low Sidelobe ConcentricHexagonal Antenna Arrays Using MOEA/D . . . . . . . . . . . . . . . . . . . . 189Sudipta Das, Durbadal Mandal, Rajib Kar and Sakti Prasad Ghoshal

QPSO for Synthesis of Scanned Linear Array Antennafor Fixed Side Lobe Level and First Null Beam WidthIncluding Wide Null Placement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197Hemant Patidar and Gautam Kumar Mahanti

Decision Feedback Equalization for MIMO Systems . . . . . . . . . . . . . . 205Samarendra Nath Sur, Rabindranath Bera and Bansibadan Maji

Fuzzy Logic-Based Handover in 3GPP LTE Network . . . . . . . . . . . . . 213Parul Datta and Sakshi Kaushal

FPGA Implementation of Novel Discrete Phase-Locked Loop . . . . . . . 223N. Bharani dharan, M. Chinnathambi and S. Rajaram

Analytical Approach on the Scale Length Modelfor Tri-material Surrounding Gate Tunnel Field-EffectTransistors (TMSG-TFETs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231P. Vanitha, G. Lakshmi Priya, N.B. Balamurugan,S. Theodore Chandra and S. Manikandan

Color Image Segmentation Using Cauchy-Mutated PSO . . . . . . . . . . . 239Suman Banerjee, Debasree Saha and Nanda Dulal Jana

Large-Scale Global Optimization Using DynamicPopulation-Based DE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251Seema Chauhan, Suman Banerjee and Nanda Dulal Jana

A Novel Preprocessing Approach for Human Face RecognitionInvariant to Illumination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261U.K. Jaliya and J.M. Rathod

Modeling and Simulation of Semi-active Jerk-Driven Damperon Severe Speed Breaker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267C.K. Nirala and S. Mandal

Voltage-Controlled Ring Oscillator for Harmonic FrequencyGeneration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277Surajit Mal, Ashis Kumar Mal and Sumalya Ghosh

Contents xiii

Amplifier Design Optimization in CMOS . . . . . . . . . . . . . . . . . . . . . . 287Sumalya Ghosh, Ashis Kumar Mal and Surajit Mal

An Application of ANFIS-Based Intelligence Techniquefor Predicting Tool Wear in Milling . . . . . . . . . . . . . . . . . . . . . . . . . . 299Shibendu Shekhar Roy

A High-performance Elliptic Curve Cryptographic Processorfor FPGA Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307V.R. Venkatasubramani, N. Premkumar, K. Vignesh and S. Rajaram

Multi-focus and Multi-exposure based Color Image Fusionfor Concealed Weapon Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313Ekta M. Upadhyay and N.K. Rana

A Compact Analytical Model for 2D Triple MaterialSurrounding Gate Nanowire Tunnel Field Effect Transistors . . . . . . . . 325D. Saraswathi, N.B. Balamurugan, G. Lakshmi Priyaand S. Manikandan

Application of Particle Swarm Optimization Techniquein Hexagonal and Concentric Hexagonal Antenna Arrayfor Side Lobe Level Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333Rajesh Bera, Durbadal Mandal, Rajib Kar and Sakti Prasad Ghoshal

Bearing-Only Tracking Using Sparse-Grid Gauss–Hermite Filter. . . . . 349Rahul Radhakrishnan, Shovan Bhaumik, Nutan Kumar Tomarand Abhinoy Kumar Singh

Novel Design and Implementation of Passive Infrared Sensorin Steel Industry Automation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357Basava Naga Girish Koneru, Vijaya Krishna Teja Mantripragada,Prithvi Raj Pani and G. Lakshminarayanan

A Comparative Study of Feature Selection and MachineLearning Methods for Sentiment Classification on Movie Data Set. . . . 367C. Selvi, Chakshu Ahuja and E. Sivasankar

Enhancement of Transmission System Loadability DuringContingency by Optimal Allocation of FACTS DevicesUsing Particle Swarm Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . 381P. Malathy, A. Shunmugalatha and P. Thaineesh

xiv Contents

Fault Diagnosis of Broken Rotor Bars in Induction MotorUsing Multiscale Entropy and Backpropagation Neural Network. . . . . 393Alok Verma and Somnath Sarangi

Design of a Signal Sensor for Analyzing Biological Activitiesat Cellular Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405Amit Krishna Dwivedi, Anubhav Sinha and Aminul Islam

Fuzzy PID Control for Ball and Beam Mechanism . . . . . . . . . . . . . . . 413Nirmal Baran Hui and Pratima Sharma

Framework for Smart Health: Toward Connected Datafrom Big Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423P. Vignesh Raja, E. Sivasankar and R. Pitchiah

Capacity Analysis of ZF Receiver in Correlated MIMONakagami-m Channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435Samarendra Nath Sur, Soumyasree Bera, Rabindranath Beraand Bansibadan Maji

Neural Estimations of First Null Beamwidth for Broadsideand End-Fire Uniform Linear Antenna Arrays . . . . . . . . . . . . . . . . . . 445Subhash Mishra, Ram Narayan Yadav and Rajendra Prasad Singh

Representing the Extended H∞ Filter (EHF) as a ModifiedExtended Kalman Filter (EKF) for Predicting the Rangeof the Robustness Bound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453Manika Saha, Ratna Ghosh and Bhaswati Goswami

Performance Evaluation of Forward Error CorrectionSchemes in Wireless Body Area Networks . . . . . . . . . . . . . . . . . . . . . 469K.S. Deepak and A.V. Babu

Simulating Self-Recovering Electric Circuits UsingNeural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479Arijit Ray, S.S. Dash and N. Chellammal

Dynamic Scenarios in Embedded System Verification . . . . . . . . . . . . . 493Hara Gopal Mani Pakala, P.L.H. Varaprasad, K.V.S.V.N. Rajuand Ibrahim Khan

Contents xv

PSO in Concentric Circular Arrays for Side Lobe Reductionwith Symmetric Relocation Boundary Condition . . . . . . . . . . . . . . . . . 503Pragnan Chakravorty and Durbadal Mandal

Ear Recognition Using Force Field Transformand Collaborative Representation-Based Classificationwith Single Training Sample Per Class . . . . . . . . . . . . . . . . . . . . . . . . 511Sayan Banerjee and Amitava Chatterjee

Demonstration of GPGPU-Accelerated ComputationalFluid Dynamic Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519Samarth Agrawal, Manish Kumar and Somnath Roy

High-performance Current Mode Receiver Designfor On-chip VLSI Interconnects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527Yash Agrawal, Rajeevan Chandel and Rohit Dhiman

CMOS Amplifier Design Using Simplified gm/ID Technique . . . . . . . . . 537Agnish Mal, Ashis Kumar Mal and Sanjit Kumar Datta

Solution of Optimal Power Flow by an Opposition-BasedGravitational Search Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545Binod Shaw, Abhik Banerjee, V. Mukherjee and S.P. Ghoshal

Information Systems and Artificial Intelligence TechnologyApplied in Concrete Road Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 559Erhan Burak Pancar and Muhammet Vefa Akpınar

Multi-hop Secured Video Transmission Using Dual-channelDual-path Routing Against Packet Copy Attack . . . . . . . . . . . . . . . . . 569Vaibhav Mishra, Saurabh Maheshwari and Kavita Choudhary

Static Analysis: A Survey of Techniques and Tools . . . . . . . . . . . . . . . 581Anjana Gosain and Ganga Sharma

A LabVIEW-Based Data Acquisition System in a QuarterCar Test Rig to Optimize Vehicle Suspension System . . . . . . . . . . . . . 593Anirban C. Mitra and Nilotpal Banerjee

A Computational Structural Biology of SoxR and DNA:A Modelling and Interactive Discern to Express the SoxOperon in Pseudaminobacter salicylatoxidans (KCT001)for Global Sulphur Oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603Sujay Ray, Arundhati Banerjee and Angshuman Bagchi

xvi Contents

Empirical Evaluations Using Character and Word N-Gramson Authorship Attribution for Telugu Text . . . . . . . . . . . . . . . . . . . . . 613S. Nagaprasad, T. Raghunadha Reddy, P. Vijayapal Reddy,A. Vinaya Babu and B. VishnuVardhan

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625

Contents xvii

About the Editors

Durbadal Mandal is Assistant Professor in the Department of Electronics andCommunication, National Institute of Technology, Durgapur, India. He completedhis B.E. in Electronics and Communication Engineering, from Regional Engi-neering College, Durgapur in 1996. He received the M.Tech. and Ph.D. degreesfrom National Institute of Technology, Durgapur, in 2008 and 2011, respectively.His research interest areas include array antenna design and filter optimization viaevolutionary computing techniques. He has over 240 research publications inJournals and Conferences of international repute.

Rajib Kar received his B.E. degree in Electronics and Communication Engi-neering, from Regional Engineering College, Durgapur, West Bengal, India in theyear 2001. He received the M.Tech. and Ph.D. degrees from National Institute ofTechnology, Durgapur, West Bengal, India in the year 2008 and 2011, respectively.Presently, he is working as Assistant Professor of electronics and communicationengineering at National Institute of Technology, Durgapur. His research interestarea include VLSI signal processing, filter optimization via evolutionary computingtechniques. He has supervised two Ph.D. students. He has published over 250research papers in International Journals and Conferences.

Dr. Swagatam Das received the B.E. Tel. E., M.E. Tel. E (Control Engineeringspecialization) and Ph.D. degrees, all from Jadavpur University, India, in 2003,2005, and 2009, respectively. Currently, he is serving as Assistant Professor at theElectronics and Communication Sciences Unit of Indian Statistical Institute,Kolkata. His research interests include evolutionary computing, pattern recognition,multi-agent systems, and wireless communication. Dr. Das has published oneresearch monograph, one edited volume, and more than 150 research articles inpeer-reviewed journals and international conferences. He is the founding co-editor-in-chief of “Swarm and Evolutionary Computation”, an international journal fromElsevier. He serves as associate editor of the IEEE Trans. on Systems, Man, andCybernetics: Systems and Information Sciences (Elsevier). He is an editorial boardmember of Progress in Artificial Intelligence (Springer), Mathematical Problems inEngineering, International Journal of Artificial Intelligence and Soft Computing,

xix

and International Journal of Adaptive and Autonomous Communication Systems.He is the recipient of the 2012 Young Engineer Award from the Indian NationalAcademy of Engineering (INAE).

Dr. Bijaya Ketan Panigrahi is Associate Professor of Electrical and ElectronicsEngineering Department in Indian Institute of Technology Delhi, India. He receivedhis Ph.D. degree from Sambalpur University. He is serving as chief editor in theInternational Journal of Power and Energy Conversion. His interests focus onpower quality, FACTS devices, power system protection, and AI application topower system.

xx About the Editors

Characterization of Fuzzy Tree Searches:A Perspective Note

Parag Bhalchandra, Santosh Khamitkar, Nilesh Deshmukh,Sakharam Lokhande and Satish Mekewad

Abstract The objective of this paper was to deal with the imprecise data involvedin different kinds of searching techniques in a more efficient way and thus tocharacterize an innovative version of search under the uncertainty. This paper is oneof the outcomes of a rigorously reviewed experimental work carried out on thesynthesis of constraint satisfaction problems through tree searching algorithmswhere it was found that the traditional tree search algorithms fail to search inuncertain problem domain.

Keywords Tree searching � Complexity � Decision trees � Fuzzy trees

1 Introduction

The real-world data analysis including data mining and classification problemsusually involves a large number of candidate inputs or features, which are almostnonlinear and are difficult enough to model [1]. Besides nonlinearity, sometimes itis difficult to choose the most relevant input variables. Thus, input selection is acrucial step with the aim of reducing the model’s complexity and removing inputs

P. Bhalchandra (&) � S. Khamitkar � N. Deshmukh � S. Lokhande � S. MekewadSchool of Computational Sciences, S.R.T.M. University, Nanded 431606, Maharashtra, Indiae-mail: srtmun.parag@gmail.com

S. Khamitkare-mail: s.khamitkar@gmail.com

N. Deshmukhe-mail: nileshkd@yahoo.com

S. Lokhandee-mail: sana_lokahande@rediffmail.com

S. Mekewade-mail: satishmekewad@gmail.com

© Springer India 2015D. Mandal et al. (eds.), Intelligent Computing and Applications,Advances in Intelligent Systems and Computing 343,DOI 10.1007/978-81-322-2268-2_1

1

that do not contribute to the model performance. Input selection methods fornonlinear systems have been studied in many places [1]. These methods can gen-erally be divided in two main groups: model-free methods and model-basedmethods [2]. Usually, these methods are straightforward and allow us to test all thesubsets of variables. While testing, we observe that there are two differentapproaches, namely top-down tree search and bottom-up tree search. Both methodsperform input selection on applications such as classification and nonlineardynamic modeling. The model obtained at every iteration is constructed usinguncertainty modeling, where the rules are identified initially. In today’s era,acquisition of knowledge is a big issue that can be improved by innovating gooddata structures so as to reduce the search complexity of traditional way of searchingfor acquisition of knowledge [2]. Approaches such as symbolic decision trees fordecision making and classification exist today with their inference procedures. Thisis very useful for people who are interested in knowledge domain understanding,classification, or the symbolic rules. This leads to extensive research effortsresulting in a number of methodological and empirical advancements [3]. Decisiontrees were proposed by Quinlan [4–6] with the ID3 program. These use symbolicdomains and hence are not useful for numeric decisions.

2 Review of Existing Approaches

In light of above discussions, we feel that the traditional trees need to be modifiedso as to reduce the tree search complexity values. This is possible only if we insistfor intelligent data analysis. To do so, we need to modify existing terminology andstructure of trees. This results in new types of trees such as decision trees, fuzzytrees, and fuzzy decision trees [1, 7]. For efficient searching in trees, we need tomodify the structural ambience and propose handling of uncertainties. This can beaccomplished by using decision trees as models and fuzzy mechanism as navigationmechanism. Many areas such as subset selection, classification, and regressionanalysis make use of decision trees. These trees use graphical model and text rulesfor navigation [7]. Since they are easy to understand and navigate, they require lesscomputational resources as compared to traditional ones. Fuzzy systems are basedon fuzzy logic and fuzzy sets, which allow us to deal with uncertain and fuzzy data.The fuzzy decision models are hybrid of fuzzy systems and decision trees. These aresuitable for providing simple, accurate, and highly interpretable models at lowcomputational costs [7]. As proposed by Prof. Zadeh [8], a fuzzy system has at leastone of its attributes defined by fuzzy logic, fuzzy set theory, and fuzzy sets. A fuzzysystem is usually composed of a knowledge base and an inference mechanism [8].The knowledge base contains a fuzzy rule base and a fuzzy data base [8]. Theformer has the rules that form the core of the system [8, 9]. These rules areconstructed based on the fuzzy sets defining the attributes of the system, stored inthe fuzzy data base which along with inference mechanism helps to classify newexamples. Fuzzy decision trees have also been proposed in the literature [3, 9, 10].

2 P. Bhalchandra et al.

There exist many algorithms to construct decision trees. One of the most usedalgorithms for constructing decision trees has long been the ID3 method introducedby Quinlan [4]. This algorithm tries to construct the smallest classification treebased on a small set of training examples. The main disadvantage is that thisalgorithm only considers symbolic discrete values both for attributes and for theresulting class. Another major disadvantage of this method includes the fact that thepre-partitioning requires previous knowledge of the data, that is, how many inter-vals are required and how to choose the split points between the intervals. Sub-sequent improvements of the ID3 algorithm, such as C4.5 [7] and CART [1, 9], tryto deal with intervals in an optimal manner as they always try to split a compactinterval into two subintervals, and they accomplish this by choosing the optimalsplit point. In terms of interpretability and ability of processing uncertainty andimprecision of fuzzy systems, fuzzy trees have been evolved as powerful models forhandling uncertainty [9]. Moreover, fuzzy decision trees inherit the desirablecharacteristics of decision trees regarding their low computational induction cost, aswell as the possibility of expressing the induced models graphically and as a set ofrules [3, 8, 11].

3 Understanding Fuzzification Process

Traditionally, decision trees have two major aspects, that is, a procedure to build thesymbolic tree and a method for decision making. We will take into considerationsome of the approaches catering to both components. The neural networks can bean alternative to our approach due to ease of application and ability to provideresponses. However, their clarity level does not match [7]. This might be a problemwhen decisions need to be justified. For such situations, we need to innovate a newapproach to combine some elements of symbolic and subsymbolic approaches. Indecision trees, the resulting tree can be cut or pruned or restructured, which oftenleads to improved generalization by incorporating additional bias [11]. In fuzzyrepresentation, knowledge can be optimized at various levels: fuzzy set definitions,norms used, etc. [2, 11]. In order to understand these concepts, we need tounderstand the basic constituents first, which are attempted in below sections. Theconcept of domain in traditional machine learning systems and in fuzzy systems istotally different. The domain for attributes can be discrete in traditional systems,whereas they are a range or continuous in fuzzy systems. Some algorithms work fordiscrete values and some work for continuous ones [2, 11]. Interchanging a con-tinuous domain with a discrete one or vice versa is not very easy and needs a goodunderstanding of the knowledge domain. This is called as covering of domain, andit is usually carried out by partitioning and clustering methods [2, 11]. Whileworking for conversion process, the covering must put every domain element intoone of the covered subset. If so, the covering is called as complete. This holds truein traditional approach but false in fuzzy sets. If the covering is inconsistent, then adomain value can be found in more than one subset. This holds true in fuzzy sets.

Characterization of Fuzzy Tree Searches … 3

Once fuzzy comes in picture, we derive the subset/domains. We then label eachderived subsets. In general, a discrete domain can be unordered, ordered partially,or ordered completely as will every discretized domain [2, 9]. Any ordering can beadvantageous as it can be used to define the notion of similarity. For example,consider the attribute skin color [11], with discrete values: white, tan, and black. Inthis case, one may say that white is “more similar” to tan than to black. Suchorderings, along with similarity measures, are being used to advantage in manyalgorithms. They will also be used to advantage in fuzzy trees (Fig. 1).

To deal with such ambiguities, additional numeric components have been addedin symbolic systems [3, 11]. This is the reason due to which decision trees havebeen extended to accommodate probabilistic measures [12] and decision rules havebeen extended by flexible matching mechanisms [11].

As another alternative, we can think of fuzzy sets. The universe of disclosure[11] represents the degree to which a set belongs. This is also known as fuzzysubset. The labels for fuzzy subsets are described by fuzzy linguistic variables,which are essentially attributes called as fuzzy terms. Their domain contains lin-guistic values [11]. We can take a fuzzy term alternatively as a fuzzy set, as thefuzzy set itself is defined by its membership function [3, 10]. For example [11],consider an attribute income. It can be continuous or discrete attribute. If we usethree linguistic terms such as low, medium, and high, it becomes fuzzy. The low,medium, and high are called as domain values. The fuzzy sets being definedsimilarly overlap to reflect fuzziness of the concepts and should cover the wholeuniverse of discourse for completeness [11]. This is illustrated in Fig. 2.

Figure 2 above shows the possibilities for actual income. As understood, the actualincome can be low, medium, and high. But the income cannot be discrete all the

Fig. 1 Understanding fuzzy trees for the skin color example

Fig. 2 Fuzzification of the income

4 P. Bhalchandra et al.

times. There can be fuzzification actual income as low and medium both with dif-ferent degrees. Fuzzy sets are generally described with convex functions [11] peakingat 1 (normal fuzzy sets). Trapezoidal (or triangular) sets are most widely used [11], inwhich case a normal set can be uniquely described by the four (three) corners valuesof U. For example, low in Fig. 2 could be represented as <0,0,10,000,25,000>.

In computer storage affiance, trapezoidal sets are very popular [7, 11]. Fuzzy setsare capable of defining classical set operations including union, intersection, andcomplements. The characteristic functions can be used for defining such operations.However, due to the infinite number of membership values, infinitely many inter-pretations can be assigned to those operations [11]. Fuzzy propositions work similarto classical propositions. We can combine them with logical connectives to formfuzzy statements. Two fuzzy sets intersect with two linguistic values if both of thefuzzy variables share the same universe of discourse. In other situation, the rela-tionship expressed in the statement can be represented with a fuzzy relation [11].The remarkable observation about fuzzy relations can be made that fuzzy setssatisfying the relation can be induced from fuzzy sets satisfying other elements ofthe relation by means of composition [7, 11, 12], provided that the relation isdefined at least once. This helps us to express the knowledge with similar rules [11].

We will address such interpretation in decision and fuzzy trees. We will stick tothe narration made in contemporary works [13]. A simple way to explain how is thecharacterization of the rules that are inferred from a decision tree is to consider allthe leaves of the tree. For each leaf, a conjunction can be easily constructed byreading the labels of the nodes and branches that are encountered starting from theroot of the tree down to the corresponding leaf. Then, in order to get the conditionfor one of the classes, we would normally construct a disjunction between all theleaves that have a value associated with that class. One of the major disadvantagesof classical decision trees is that each tested object will have only one leaf asso-ciated with it. Moreover, this leaf is closely related to the input samples. In manycases, it would be a nice feature to have close values somehow related to each other.Also, it would be nice to be able to construct decision trees by using symbolicvalues such as tall and short, and hot and cold. All these issues are solvable using afuzzy approach [4, 5]. Going back to the classical decisions inferred by a decisiontree, in order to test an object against a given decision tree, we would start from theroot of the tree and go down the tree by testing the attribute corresponding to thecurrent node and following only the branch of the tree that corresponds to the valuethat has the same value as the object we are testing [6, 14]. In fuzzy terms, thiswould mean to follow only the branch corresponding to the set where our value hasa strictly positive membership degree. In the case of crisp decision trees, only onebranch will have degree 1 and all the other branches will have degree 0, but for ourfuzzy approach, we will consider all the branches for which the membership degreeof our object to that particular set is nonzero. For each node, we will also have tokeep in mind the membership of the object down the current path from the root ofthe node, where the membership is always 1, down to the current node. Again, inthe case of crisp decision trees, the membership on the current path is always 1(only one path and only one leaf have membership 1, all the other ones have

Characterization of Fuzzy Tree Searches … 5

membership 0 and are therefore ignored). For simplicity, we will refer to themembership degree of an object to a certain set of attributes as the membership ofthe object to the corresponding tree node (or to the corresponding path in the tree).This membership can be computed gradually and needs to use an intersectionoperator for fuzzy memberships. In other words, considering that the attributesA1, A2,. . .Ak were encountered from the root of the tree to a certain node (the nodewill in this case be on level k), below table shows the comparison between crisp andfuzzy decision trees [8, 9, 15] (Fig. 3).

The algorithm complexity of constructing a kind of minimum fuzzy decision treeis NP-hard. Construction of such trees needs considering a directed tree whose eachedge links two nodes, the initial node, and the terminal node. The former is calledthe father node of the latter, while the latter is said to be the son node of the former.The node having not its father node is the root, whereas the nodes having not anyson nodes are called leaves. The general matching strategy of fuzzy decision trees isdescribed as follows:

1. Matching starts from the root and ends at a leaf along the branch of the max-imum membership.

2. If the maximum membership at the node is not unique, matching proceeds alongseveral branches.

3. The decision with maximum degree of truth is assigned to the matching result.

We illustrate this with an example. Consider a dataset [15] as in Fig. 4).In the above figure, each column corresponds to a fuzzy subset defined as

X = {1, 2, 3, …, 16}. In the last column, three symbols V, S, and W denote threesports to play on weekends: volleyball, swimming, and weightlifting, respectively.Now, keeping in mind the construction process as explained in previous pages, wedefine four attributes as follows:

Fig. 3 Comparison of crisp and fuzzy decision trees

6 P. Bhalchandra et al.

Temperature ¼ hot; mild; coolf g � FðXÞOutlook ¼ sunny; cloudy; rainf g � FðXÞ

Humidity ¼ humid; normalf g � FðXÞWing ¼ windy; not windyf g � FðXÞ

A fuzzy decision tree for the root Temperature is given in Fig. 5. The percentageattached to each decision is the degree of truth of the decision. Consider twoexamples remaining to be classified

Fig. 4 Illustrative dataset [15]

Fig. 5 The process of matching in the fuzzy decision tree for temperature example

Characterization of Fuzzy Tree Searches … 7

e1 ¼ 0:0; 0:6; 0:4; 0:3; 0:7; 0:0; 0:5; 0:5; 0:5; 0:5ð Þ ande2 ¼ 0:9; 0:1; 0:0; 0:8; 0:2; 0:0; 0:5; 0:5; 0:8; 0:2ð Þ:

The process of matching in the fuzzy decision tree is shown in Fig. 5.Keeping all the above discussions in mind, we can refine the traditional tree

searchers. Prof. Ashit Kumar Dutta [13] has already highlighted these consider-ations. He has considered a simple search tree with one root node/start node andsome descendent nodes. Consider a simple search path from root node/start node tosome node in the tree. Accordingly, each node is labeled with only the terminalnode of the path it denotes. Each child denotes a path that is a one-city extension ofthe path denoted by its parent. The most important searching method [16–18] thathave analyzed so far includes DFS, BFS, Hill Climbing, A*, and IDA*. The depth-first search is a good idea if depth is less and breadth is more. In contrast, depth-firstsearch is a bad idea if there are long paths, even infinitely long paths, that neitherreach dead ends nor become complete paths. In those situations, we need alternativesearch methods. Breadth-first search works even in trees that are infinitely deep oreffectively infinitely deep. On the other hand, breadth-first search is wasteful whenall paths lead to the goal node at more or less the same depth. The breadth-firstsearch is a bad idea if the branching factor is large or infinite, because of expo-nential explosion. Breadth-first search is a good idea when we are confident that thebranching factor is small. We may also choose breadth-first search, instead ofdepth-first search, if we are worried that there may be long paths, even infinitelylong paths, that neither reach dead ends nor become complete paths [13, 18]. Wemay be so uninformed about the search problem that we cannot rule out either alarge branching factor or long useless paths. In such situations, we may want toseek a middle ground between depth-first search and breadth-first search. Searchefficiency may improve spectacularly if there is a way to order the choices so thatthe most promising ones are explored at the earliest. In many situations, we canmake measurements to determine a reasonable ordering. We have also learnedabout such methods that take advantage of such measurements; they are calledheuristically informed methods such as Hill Climbing, A*, and IDA*. For bettersearch efficiency, we can use the concept of shortest path. By the shortest path, wemean the path of which total distance is minimum [17, 18]. An obvious procedure isthat one can find out all possible paths (if possible) and select the best one fromthem. For this, one can find out all possible paths by DFS or BFS and the search hasto be continued until all paths are found. But in the maximum case, the site of thesearch tree is too large and consequently finding all possible paths is extremelyrigorous. In many cases, we can improve the existing searching technique by usingguesses about distances remaining, as well as facts about distances already accu-mulated. After all, if guess about distance remaining is good, then that guesseddistance added to the definitely known distance already traversed should be a goodestimate of total path length. Surely, it makes sense to work hardest on developingthe path with the shortest estimated path length until the estimate is revised upwardenough to make some other path be the one with the shortest estimated path length

8 P. Bhalchandra et al.

[6, 13, 15]. After all, if the guesses were perfect, this approach would keep us on theoptimal path at all times. In general, however, guesses are not perfect, and a badoverestimate somewhere along the true optimal path may cause us to wander awayfrom that optimal path permanently. Here, we make an interesting note that theunderestimates cannot cause the right path to be overlooked. An underestimate ofthe distance remaining yields and underestimate of total path length u. Now, if wefind a total path by extending the path with the smallest underestimate repeatedly,we need to do no further work once all partial path distance estimates are longerthan the best complete far encountered [13]. We can stop because the real distancealong a complete path cannot be less than an underestimate of that distance if allestimates of remaining distance can be guaranteed to be underestimates. Whiledoing so, we need to remember that the closer an underestimate is to the truedistance, the more efficiently we search. It is because, if there is no difference at all,there is no chance of developing any false movement. At the other extreme, anunderestimate may be so poor as to be hardly better than a guess of zero, whichcertainly must always be the ultimate underestimate of remaining distance [6, 13,15]. In fact, ignoring estimates of remaining distance altogether can be viewed asthe special case in which the underestimate used is uniformly zero.

4 Conclusion

This paper highlights a method of interpreting decision tree searches by using fuzzylogic. This interpretation is labeled as characterization of fuzzy tree searches.Although a lot of issues are still up for discussion such as choosing the appropriatefuzzy operators and choosing the crisp partition on each attribute. By using thisfuzzy characterization of searches, we get a new interpretation of the nodes, whichhelp us to achieve greater accuracy when computing in uncertainties.

Acknowledgments Prof. Ashit Kumar Dutta, Prof. Richard Korf.

References

1. Mendonça, L.F., Susana M.V., Sousa, J.M.C.: Decision tree search methods in fuzzy modelingand classification. Int. J. Approximate Reasoning 44(2), 106–123 (2007)

2. Pol, C.: Fuzzy decision trees. Lecture Notes. Available at www.cs.ubbcluj.ro/*gabis/DocDiplome/DT/FuzzyDT/fuzzy

3. Zadeh, L.A.: A theory of approximate reasoning. In: Hayes, J., Michie, D., Mikulich, L.I.(eds.) Machine Intelligence, vol. 9, pp. 149–194. Halstead Press, New York (1979)

4. Quinlan, J.R.: Induction on decision trees. Mach. Learn. 1:81–106 (1986)5. Quinlan, J.R.: Decision trees as probabilistic classifiers. In: Proceedings of the Fourth,

International Workshop on Machine Learning, pp. 31–37 (1987)6. Kahraman, C.: Investment Analyses Using Fuzzy Decision Trees. Lecture Notes. Istanbul

Technical University, Department of Industrial Engineering, 34367, Macka, İstanbul, Turke

Characterization of Fuzzy Tree Searches … 9

7. Zadeh, L.A.: Fuzzy sets. Inf. Control 8, 338–353 (1965)8. Zadeh, L.A.: Fuzzy logic and approximate reasoning. Synthese 30, 407–428 (1975)9. Cintra, M.E., Monard, M.C., Camargo, H.A.: Fuzzy DT—a fuzzy decision tree algorithm

based on C4.5. In: National or Ibero-american conference publications-CBSF’2012: CongresoBrasileiro de Sistemas Fuzzty, Natal, RN, Brasil (2012)

10. Jager, R.: Fuzzy logic in control. Ph.D. dissertation, Technische Universiteit, Delft (1995)11. Janikow, C.Z.: Fuzzy decision trees: issues and methods. Department of Mathematics and

Computer Science, University of Missouri-St. Louis12. Jang, J.: Structure determination in fuzzy modeling: a fuzzy CART approach. In: Proceedings

of the IEEE Conference on Fuzzy Systems, pp. 480–485 (1994)13. Dutta, A.K.: Intutionistic fuzzy search using dynamic programming. J. Comput. Sci. Inform.

Technol. 3(1), 1–6 (2006). ISSN 0973-4872 (Institute of Technology and Management,Gurgaon, Voyager)

14. Quinlan, J.R.: The effect of noise on concept learning. In: Michalski, R., Carbonell, J.,Mitchell, T. (eds) Machine Learning II. Morgan Kaufmann, Burlington (1986)

15. Yuan, Y., Shaw, M.J.: Induction of fuzzy decision trees. Fuzzy Sets Syst. 69, 125–139 (1995)16. Bhalchandra, P.U., Khamitkar, S.D., Lokhande, S.N.: Pathology of trees, tree searches and

their computational complexities. JIKRCSA Int. J. 2(2) (2014)17. Bhalchandra, P.U., Khamitkar, S.D., et al.: A review of contemporary research investigations

in selected brute force and heuristics tree search algorithms. IJSHRE Int. J. 2(1) (2014)18. Bhalchandra, P.U., et al.: Experimental analysis of selected constraint satisfaction problems

using tree searching algorithms. In: Proceedings of ACM Sponsored ICCI, Mumbai (2014)

10 P. Bhalchandra et al.

A Modified SOM-Based RBFNfor Rotation Invariant Clear and OccludedFingerprint Recognition

Sumana Kundu and Goutam Sarker

Abstract In this paper, a modified radial basis function network (RBFN) based onself-organization mapping (SOM) has been designed and developed for rotationinvariant clear as well as occluded fingerprint recognition. The SOM-based RBFNlearns different fingerprint images and performs subsequent rotation invariant rec-ognition of clear and occluded images. The system is efficient, effective, and fast.Also, the performance evaluation of the system is substantially moderate.

Keywords SOM � RBFN � BP learning � Occlusion � Holdout method �Accuracy � Precision � Recall � F-score

1 Introduction

Fingerprint recognition is one of the vital biometrics methods that has been widelyused in various applications because of it is reliability and accuracy in the processof recognizing and verifying a person identity.

Most of the fingerprint recognition methods are based on feature (minutiae)extraction and minutiae matching [1, 2]. In [3], a fingerprint identification tech-nology was represented by minutiae feature extraction using back propagation (BP)algorithm.

A three-rate hybrid Kohonen neural network (TRHKNN) was proposed in [4]for distorted fingerprint image processing in conditions of wide variation in degreeof distortion. This proposed TRHKNN is capable to identify the distorted image offingerprint and restore the undistorted image of fingerprint.

S. Kundu (&) � G. SarkerComputer Science and Engineering Department, NIT, Durgapur, Indiae-mail: sumana.kundu@yahoo.co.in

G. Sarkere-mail: g_sarker@ieee.org

© Springer India 2015D. Mandal et al. (eds.), Intelligent Computing and Applications,Advances in Intelligent Systems and Computing 343,DOI 10.1007/978-81-322-2268-2_2

11