A Simple Kernel Method with Shannon Wavelet Kernel for Face Recognition

To avoid using all training face images to recognize a test face image, the class centroids can be used; to alleviate the impairment of recognition accuracy due to face occlusion, Shannon wavelet kernel can be utilized. This paper presents a straight forward kernelbased approach with Shannon wavelet kernel for face recognition, which first finds the centroid face image in the kernel feature space for each class, and then compares test images with every kernel class centroid to decide which they belong to. Theoretical analysis and experimental results on Yale face database both show that our method is simple but capable of fast face recognition while maintaining high recognition accuracy.