Identification of a Person Using Multimodal Biometric System

Abstract

A biometric identification system is an automatic patternrecognition system that recognizes a person by determiningthe authenticity of a specific physiological and/or behavioralcharacteristic (biometric) possessed by that person.Unimodal biometric systems often face significantlimitations due to sensitivity to noise, intraclass variability,data quality, nonuniversality, and other factors. To improvethe performance of individual matchers in such situationsmay not prove to be highly effective. Multibiometric systemsseek to alleviate some of these problems by providingmultiple pieces of evidence of the same identity. Thesesystems help achieve an increase in performance that maynot be possible using a single-biometric indicator. Aneffective fusion scheme that combines information presentedby multiple domain experts based on the rank-level fusionintegration method. The developed multimodal biometricsystem possesses a number of unique qualities, starting fromutilizing principal component analysis and Fisher’s lineardiscriminant methods for individual matchers (face, ear, andsignature) identity authentication and utilizing the novelrank-level fusion method in order to consolidate the resultsobtained from different bio- metric matchers. The ranks ofindividual matchers are combined using the highest rank,Borda count, and logistic regression approaches. The resultsindicate that fusion of individual modalities can improve theoverall performance of the biometric system, even in thepresence of low quality data. Moreover better performancecan be obtained by using ICP algorithm for ear database.