Iris verification using wavelet moments and neural network

In this paper, a novel and robust verification approach using iris features is presented. Contrasting with conventional approaches, only two iris subregions instead of entire iris, where are nearly not occluded by useless parts such as eyelash and eyelid, are segmented for verification. Gabor filtering and wavelet moments methods are used to extract the iris texture features. In the verification stage, the principal component analysis (PCA) technique and one-class-one-network (Back-Propagation Neural Network (BPNN)) classification structure are employed for dimensionality reduction and classification, respectively. The experimental results show that the correct verification rate can reach 98.65% using our proposed approach.