Dorsal hand veins biometrics using NIR images with fusion of classifiers at score level

Abstract

This paper presents a biometric system on dorsal hand vein images in the near infrared (NIR), with an approach based on fusion of classifiers at score level. Fiducial features containing information on texture and shape are used with two classifiers based on Chi-square distance and Dynamic Time Warping (DTW), respectively, and further fused at score level. A collection of experiments using a publicly available dataset obtained from Universidad de Las Palmas de Gran Canaria was carried out. The obtained results indicate an Equal Error Rate of EER=0.0486 and EER=0.0274 and in average with classifiers fusion using sum and multiplication of scores in verification mode, and recognition rate of RR=95.80% and RR=97.30% in identification mode, respectively. These results represent an improvement with respect to results obtained when both classifiers and features are used individually.