Iris Recogniton Based on Fearure Extraction in Kernel Space

Iris-based recognition approach is a popular and efficient method in personal identification field. How to code an iris image is the key issue for iris recognition. In this paper, we will apply Kernel-based nonlinear feature extraction Kernel Principal Component Analysis (KPCA), Kernel Independent Component Analysis (KICA), Kernel Linear Discriminant Analysis (KLDA), and Kernel Springy Discriminant Analysis (KSDA) to encode an iris image. The idea of Kernel-based feature extraction methods is to map the input data into an implicit feature space F with the kernel trick firstly, and then perform original linear methods to produce nonlinear projection matrix of input data. The performances of these encoding methods are analyzed using CASIAII database. We plot a series of Receiver Operating Characteristics (ROCs) and Equal Error Rate (EER) to demonstrate various the different performances of different encoding methods.