A New Method for Face Recognition

For the wireless terminals with the low memory and limited computing performance, it is necessary to research face recognition strategies with low complexities and small memory requirements. Sparse Representation Classifier (SRC) has been focused on because it is simple and can achieve high accuracies even for down-sampling facial images. Therefore, a new method based on SRC is proposed to obtain a higher accuracy in condition of low memories and computing performance. Furthermore, the real system consisting of CMOS cameras and computers is constructed using the proposed algorithm. The experimental results are consistent to the simulation results based on ORL, AR and UMIST databases, verifying that the algorithm and the system based on the algorithm can acquire a higher accuracy and be more robust to occlusions.