Performance Evaluation of Multimodal Biometric Authentication System

Abstract

ABSTRACT newlineBiometrics is the area of automatic identification of persons based on the physical and/or newlinebehavioral characteristics of the human body. The utilization of biometrics in the field of newlineimproving security and verification in delicate systems is a quickly creating innovation. newlineThe unimodal biometric system has numerous confinements about precision, execution, newlinesecurity, and toughness. The decreased security and increased spoof attacks in unimodal newlinesystems have resulted in the development of a novel system known as multimodal systems newlineformed by uniting different biometric characters. A multimodal biometric system offers a newlinesolution to these problems by combining information from multiple sources and provides newlinebetter recognition performance as compared to the unimodal system. Multimodal systems newlinereduce failure to enroll and resist spoofing as multiple biometric sources cannot be spoofed newlinesimultaneously. Multimodal biometrics is the level based approach where fusion takes place newlineat different levels as sensor, feature, matching score and decision. Fusion at matching newlinethe score level provides better recognition performance as it contains more contented newlineinformation which is both feasible and practical. newlineThis research work aims to analyze the performance of three biometric traits, such as newlinefingerprint, face, and iris, separately and combine them using matching score level fusion newlinerule and the classification is performed using three soft computing approaches, namely newlineSVM classifier, sparse SVM classifier and Rough set fuzzy classifier. The matching score newlinelevel fusion rule is optimized using Adaptive Particle Swarm Optimization(PSO) to ensure newlinethe desired system performance corresponding to the desired level of security. The face, newlineiris and finger print images in the database are first preprocessed using the median filter newlineto remove any noise present has extracted a region of interest, which can be used for newlinefeature extraction. This step serves to improve the quality of the images and to extract only newlineuseful information and also reduce