Support vector machine with local summation kernel for robust face recognition

This paper presents support vector machine (SVM) with local summation kernel for robust face recognition. In recent years, the effectiveness of SVM and local features is reported. However, conventional methods apply one kernel to global features. The effectiveness of local features is not used in those methods. In order to use the effectiveness of local features in SVM, one kernel is applied to local features. It is necessary to compute one kernel value from local kernels in order to use the local kernels in SVM. In this paper, the summation of local kernels is used because it is robust to occlusion. The robustness of the proposed method under partial occlusion is shown by the experiments using the occluded face images. In addition, the proposed method is compared with the global kernel based SVM. The recognition rate of the proposed method is over 80% under large occlusion, while the recognition rate of the SVM with global Gaussian kernel decreases dramatically.