Biometric Fusion Using Enhanced SVM Classification

Support Vector Machines or SVM is one of the most successful and powerful statistical learning classification techniques. It has been also implemented in the biometric field. In this paper we propose the use of SVM as a fusion tool. We propose a system that fuses the classification obtained from the iris biometric and the fingerprint biometric. In addition, we show how score normalization can have a dramatic effect on performance (and the speed). The proposed model leads to considerable improvement in accuracy. In fact, the new fusion model improved the classification accuracy from around 96% for the best single biometric (the iris in this case) to over 99.8%. We believe that fingerprint and iris are a good combination to fuse and hope that this merits further research.