A Face Recognition Method Based on Singular Value Disturbance, Algorithm of SLLE, and Dynamic RBF Neural Network

A new face recognition method based on singular value disturbance, algorithm of supervised locally linear embedding (SLLE), and dynamic RBF neural network is proposed. Firstly, the new sample set was formed by the fusion of original face image in ORL face database and the image generated by singular value disturbance, which effectively extends the limited sample set. Then, the algorithm of SLLE was applied to reduce the dimension of the fused image, which increases the Dijkstra distance between classes. Finally, the dynamic RBF neural network classifier is responsible for the classification. In order to implement classification, weights and classification error should effectively and continually be readjusted. The experimental results based on ORL face database show that the performance of the recognition approach is better and real time.