Certain investigations on meta heuristic algorithms for feature selection in multimodal biometrics

Abstract

Biometrics is a technical term used for both the security of the newlinepersonal and the financial data of an individual. This technology focuses on newlinephysiological characters that are unique for an individual such as fingerprint, newlineface, iris, retina, ear, hand geometry, hand veins, palmprints, voice and newlinebehavioural traits such as gait, signature and keystroke dynamics. Fingerprint, newlineiris and face are considered for composite mode of analysis in this research newlinework. Feature extraction is used to specify an image with various features set newlinefor removing the redundancy and dimensions of the data. The feature newlineextraction module is developed on biometric data and it extracts only the newlinerepresentative material to form a new ensemble of the data. This ensemble newlineshould be unique for each person. The feature extraction was carried out by newlineusing Gabor and Local Tetra Pattern (LTrP) for different composite modes newlinelike fingerprint-iris, fingerprint-face and fingerprint-iris-face. The fingerprintiris newlineperforms better in feature extraction using Gabor filter on arelative newlineaverage GAR of 4.11% and 9.05% when compared to fingerprint-face and newlinefingerprint-iris-face respectively. In addition, the fingerprint-iris performs newlinebetter in feature extraction using LTrP on a relative average GAR of 4.63% newlineand 9.72% when compared to the combinations of fingerprint-face and newlinefingerprint-iris-face correspondingly newline newline