An Iris Recognition Approach based on Fuzzy Support Vector Machine

Abstract

An iris recognition system named IrisPassport is presented in this paper. Standard Deviation is used to localize the irises from iris images. After localization, IrisPassport uses Steerable Pyramid and Variant Fractal Dimension as features with orientation information. Aiming to build a robust solution for non-cooperative iris images, we adopt fuzzy support vector machine (FSVM) because we consider different samples contributes to classification differently and a member function can be used when unclassifiable regions appear. Experimental data demonstrates the potential of our new approach, and shows that it performs favorably when compared with the former algorithms.