A new Kernel function based face recognition algorithm

In general, using kernel function to solve problems of non-linear and recognition ratio, as is done in human face recognition, is particularly effectively. Firstly, a hybrid kernel function is constructed, and then a modified human face recognition algorithm about Kernel-based KICA and Kernel-based improved PSVM methods is presented. The traditional ICA methods have limitations for non-linear image in facial feature extraction process. Using kernel-based non-linear image characteristics, KICA method analyses data in the high-dimensional feature space. As a machine learning algorithm, SVM also has some limitations. This article presents an improved Non-linear PSVM algorithm to get a better recognition ratio and a little time consuming. Eventually the tests for feasibility are performed.