Single-modal biometrics has certain inherent problems such as noisy sensor data, non-universality of biometric traits, restricted degrees of freedom and the stability of the single algorithm itself. In order to overcome these problems and obtain a satisfactory recognition rate while maintaining high robustness, an effective solution is to employ multi-modal biometrics. Specifically, single-modal biometric multiple representation fusion is a form of multi-modal biometrics, which involves using multiple representations on a single biometric indicator. Moreover, multiple representation fusion is actually feature fusion. The key to feature fusion is how to deal with the local uncertainty. Drawing lessons from the human cognitive process, manifold, is introduced in order to realize a smooth transition from local to global on the basis of topology. In this paper, we present a novel scheme for fusing the Palmprint Mixed-Phase Features and the Palmprint Directional Valley Features, termed Shape of Gaussian (SOG) matching, which yields equivalent results to the feature fusion approaches employed in previous work. Furthermore, since SOGs form a Lie group, and Lie group is an important kind of manifold, a distance metric for SOG based on Lie group theory is adopted. Experimental results illustrate the effectiveness of the proposed approach.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below