An accurate infrared hand geometry and vein pattern based authentication system

Abstract

In this paper, hand dorsal images acquired under infrared light are used to design an accurate personal authentication system. Each of the image is segmented into palm dorsal and fingers which are subsequently used to extract palm dorsal veins and infrared hand geometry features respectively. A new quality estimation algorithm is proposed to estimate the quality of palm dorsal which assigns low values to the pixels containing hair or skin texture. Palm dorsal is enhanced using filtering. For vein extraction, information provided by the enhanced image and the vein quality is consolidated using a variational approach. The proposed vein extraction can handle the issues of hair, skin texture and variable width veins so as to extract the genuine veins accurately. Several post processing techniques are introduced in this paper for accurate feature extraction of infrared hand geometry features. Matching scores are obtained by matching palm dorsal veins and infrared hand geometry features. These are eventually fused for authentication. For performance evaluation, a database of 1500 hand images acquired from 300 different hands is created. Experimental results demonstrate the superiority of the proposed system over existing systems.