Non-linear classification for iris patterns

Biometric authentication based on iris patterns is used for personal identification. Important attributes to identity applications include accuracy, speed and template size. Iris patterns are segmented by considering the maximum area of the connected components in the binary images. The iris region is decomposed into subregions. Hu moments are applied to the minimum variance subregions (MVS). The summation of the moment values in these subregions is given as input to Support Vector Machine (SVM) and feed forward Neural Network (NN). The prominent results of False Rejection Rate (FRR) 2.5% and False Acceptance Rate (FAR) 0.0% was obtained for SVM. With NN classifier, prominent results of FRR 4.2% and FAR 0.0% are obtained.