基于支持向量机的虹膜识别方法研究

In order to improve accuracy and stability for iris recognition, this paper studied support vector machine(SVM) learning algorithm based on statistical learning theory and expanded two-dimension iris feature to multi-dimension situation for multi-dimension iris feature. Then, based on the performance of SVM in tackling small sample size, high dimension and its good generation, the paper proposed an iris recognition method based on SVM. The method realized to map iris features to high dimension space using radial based function. The CASIA iris database was used to test the proposed method. The experimental results show that comparing with the nearest feature line method and dissimilarity functions, recognition rated of the proposed method is increased 3.83% and 5.76% respectively, while improving the stability and flexibility of iris recognition for kernelization of SVM.