Feature level fusion of face and palmprint using Gaussian mixture model: application to single sample analysis

Abstract

With growing concern about the security, the world over, biometric based person verification is gaining more and more attention. One of the major limitation in Biometric authentication is single sample biometric recognition (unimodal) problem. In this paper, we combine two biometrics namely face and palmprint at feature level using the novel approach based on Log Gabor transform and Gaussian Mixture Mo- del to address this problem. The proposed technique consists of three important steps: First, we vertically fuse the texture features of face and palmprint that are extracted separately using Log-Gabor transform. Second, we model the fused texture data using Gaussian Mixture Model (GMM) to obtain more than one texture transformation matrices. Third, we analyze each of these texture transformation matrices using Principle Component Analysis (PCA) / Independent Component Analysis (ICA) separately. Extensive experiments are carried out on large face and Palmprint databases to prove the efficacy of the proposed method. The experimental results show the superiority of the proposed method compared to some of the existing schemes.