Improved method based on NMF for face recognition

A method of dealing with NMF (non-negative matrix factorization) basis to enhance face recognition rate is introduced in this paper. Firstly, we use discrete wavelet transformation to produce a representation in the low frequency domain, and get basic matrix according to the NMF method. Secondly, parts of face features which possess outstanding performance are extracted by threshold value judgments, and they are used to form optimized facial subspace feature. The training and testing images are projected to the optimized subspace feature. Finally, support vector machine is used for classification. Experiments show that the improved method is workable, especially under the circumstance of partial occlusion, it achieves remarkable effects.