User-Specific Iris Authentication Based on Feature Selection

A novel iris localization method and user-specific automatic iris authentication approach based on feature selection is proposed in this paper. First, two iris sub-regions, where are nearly not occluded by useless parts such as eyelash and eyelid, are segmented as region of interest (ROI). Second, multi-scale Gabor filters are adopted to extract the texture feature of ROI. Third, genetic algorithm (GA) and support vector machine (SVM) are employed for feature selection and classification. Through feature selection, each user has specific feature index and authentication modality. For proving the effectiveness and feasibility, we compare the proposed specific feature selection approach with the method without feature selection on a small database. The experimental results show the proposed approach can achieve lower error rates in iris authentication.