Palm vein recognition based-on minutiae feature and feature matching

Abstract

Palm vein recognition is one of the biometric systems that recently explored. The location of palm-vein that inside the human body, give a special characteristic compare with other biometric modal. It expected to be robust, difficult to be duplicated, and are not affected by dryness and roughness of skin. Therefore palm vein has high security and needs to be studied more. In this paper we develop a recognition system consists of several processes; they are ROI detection using peak-valley detection and first CHVD rules, pre-processing using maximum curvature, feature extraction based-on minutiae, and feature matching using based-on weighted Euclidean score. The experimental result yielded a best success rate of 91.00% in term of accuracy with configuration of the system using adaptive histogram equalization, full minutiae feature and the group-voting matching (the threshold for point matching set at 0.10). In term of biometric performance we achieve Equal Error Rate at 9.94% with threshold 0.50380. Both of best performance achieve with only 42 average number of minutiae feature.