Evaluation of support vector machine with universal kernel for hand-geometry based identification

Hand-geometry based authentication is gaining widespread application in a number of security systems. It can be operating in either verification or identification mode. Although the verification mode has received great attention in research in the past, the identification mode is still an open research area and new innovative solutions are needed to reduce the computational time and enhance accuracy. In this paper, we explore and evaluate a new approach based on support vector machines with universal kernel for addressing this problem. We also compare its performance with some other kernel functions and common classifiers including rule based and decision-tree based classifiers. Our experiments reveal significant improvements in the performance of hand geometry based identification for the proposed approach on the adopted dataset as compared to other approaches. More than 98% average identification accuracy can be achieved with less than 0.04% average false acceptance rate and 2% average false rejection rate.