Multibiometric template generation using CS theory and discrete wavelet transform based fusion technique

Abstract

Nowadays, two or more than two biometric traits or templates are used for individual verification by many institutions due to limitations of the unimodal biometric system. Two or more than two biometric templates are used for recognition by individuals in biometric system which is known as multibiometric system. The problem of generation of multibiometric template is associated with any multibiometric system designing. In this paper, fusion technique for generation of multibiometric template is proposed using compressive sensing theory and Discrete Wavelet Transform. In this technique, sparse measurements of fingerprint image are fused into diagonal wavelet coefficients of face image of the same individual to generate multibiometric template. The fused face image is used for individual recognition for decisions about verification. Defused sparse measurements of fingerprint image from fused face image and then reconstructed fingerprint image using CS recovery processes from extracted sparse measurements. This reconstructed fingerprint image is used for the second level of verification of individuals. This verification accuracy of face system fingerprint system and multibiometric system is above 80 % after application of the proposed fusion algorithm.