Support vector machine is a new machine-learning method developed on statistics learning theory. It is used to resolve the problems such as small-samples, nonlinear and hyper-high dimension in pattern recognition domain. And it shows outstanding advantages in treating generalization and learning performances. It is a cogent classification instrument. The face recognition based on support vector machine is one of the active research topics and hotspot of science and technology in recent years.The basic theory and system composing of face recognition were introduced firstly in this paper. Then the method of kernel independent component analysis used in face features extraction and the face recognition technology of mixture kernels function SVM were emphasized.Firstly, we introduced the face recognition technology, developmental status and system composing. Then, the method of KICA which was implemented in face features extraction were presented after the method of ICA were introduced detailed. In the end, the method of SVM and kernel function were studied particularly. To obtain the local and gloabal, generalization and learning performances quired, we constructed a new mixture kernels function which we combined with SVM method to implement face classification and recognition in accordance with the problems of multi-classification and nonlinear in face recognition.The feasibility of the algorithm was demonstrated by experiment on ORL face database. The experimental results showed that KICA used in face features extraction can better overcome the effect of light and shade as well as angles variation, also with a encouraging adaptive ability of nonlinear. The effect of the face classify and recognize methods based on mixture kernels function SVM was much better than others. |