Face Feature Selection with Binary Particle Swarm Optimization and Support Vector Machine

Hongtao Yin,Jiaqing Qiao,Ping Fu,XinYuan Xia
2014-01-01
Abstract:A face feature selection and recognition method based on BPSO and SVMWrapper model is presented. To solve the problem that DCT coefficients dimension is higher for face recognition, we design a SVM-Wrapper model based on BPSO. In the process of training SVM, the cross-validation is used to training samples, and the recognition accuracy is used for defining the fitness function of BPSO feature selection algorithm. The fitness function is used to guide the BPSO algorithm to search the optimal feature subset. The experiments on ORL databases show that the improved method is effective.
What problem does this paper attempt to address?