Study on Finger-Articular Back Texture recognition

Abstract

Humanpsilas finger-articular back texture(FABT), as a novel biometric identification pattern, has been studied. As a basis set of FABT space, eigenjoints are extracted by principle component analysis. The features of each finger-articular back texture are computed by projecting on the related eigenjoint space. In matching stage, the decision are made by using nearest neighbor classifier based on Mahalanobis distance. The results show that: back finger- joint texture has high uniqueness in terms of high recognition accuracy rate (97.57 percent); the inter-class and intra-class have good separability; and recognition speed is fast enough for real time identification.