A face recognition method based on joint sparse representation of multiple features is proposed in this paper. First, principle component analysis (PCA), kernel PCA (KPCA), and non-negative matrix factorization (NMF) are used to extract feature vectors of face images. The three features could provide complementary descriptions for face images. Then, in the classification stage, joint sparse representation is employed to classify the three features thus considering their correlations. Finally, the total reconstruction errors of the three features on different kinds of training classes are calculated to determine the label of test sample. Experiments are conducted on AR and Yale-B databases to validate the effectiveness of the proposed method.
展开▼
