Face recognition base on KPCA with polynomial kernels

Kernel principal component analysis (KPCA), a improving of PCA, is used in face recognition. The paper describes the use of kernel principal component analysis with polynomial kernels to extracts face image features in high-dimensional spaces. KPCA extracts feature set more suitable for categorization than classical Principal Component Analysis does. The experiments on the ORL and Yale face database demonstrate that KPCA is good at dimensional reduction, and it achieves better performance than classical Principal Component Analysis does, the highest correct recognition rate is 99%.