Multimodal Biometric Identification system using Random Selection of Biometrics

Abstract

Biometric systems employ their biometric features to identify people. Identification systems that solely employ one biometric modality would not be able to meet the demands of demanding biometric applications in terms of performance, acceptance, and uniqueness. The majority of unimodal biometrics systems have problems with concentrated data noise, variances within and across classes, non-universality, etc. Multimodal biometric systems, which may establish identity from many sources of information, can bypass some of these restrictions. Identifying a person using multimodal biometric technology is more accurate and dependable. Early integration tactics are anticipated to perform better than late integration strategies. In this paper, feature-level fusion using the random selection of biometrics is presented. Block variance features and contourlet transform features are used to carry out the feature-level fusion. LDA is used to reduce the feature vector's dimensions. When compared to alternative integration approaches and their unimodal cousin, integrating the contourlet transform features of two independently determined biometric qualities delivers a consistent gain in performance accuracy. In this work, we use a random selection of biometric traits to guarantee the presence of a real human being at the time of data collection. Only fingerprints, palm prints, and faces will be included in the random selection.