An Effective Iris Recognition System Based on Combined Feature Extraction and Enhanced Support Vector Machine Classifier

Abstract

An effective iris recognition system is proposed in this study. Firstly, both Gray Level Co-occurrence Matrix (GLCM) and multi-channel 2D Gabor filters are adopted to extract iris features. GLCM is a statistic method. 2D Gabor filters reflect features of spatial and frequency transformation. The combined features are in the form of complementary and efficient effect. Secondly, Particle Swarm Optimization (PSO) is employed to deal with the parameter optimization for Support Vector Machine (SVM), and then the optimized SVM is applied to classify iris features. The experimental results demonstrate that our proposed iris recognition system outperforms some of the existing methods. And the SVM optimized by PSO achieves higher recognition accuracy and lower standard deviation than that of the SVM using grid search method. The recognition rate of 99.409% obtained on JLUBRIRIS-V1 iris image database indicates the proposed iris recognition system has great potential for practical use.