Research on Correction Model of PSVM in Face Recognition

Proximal support vector machine (PSVM) has the advantage of short training time, however, it has not high recognition rate because of correction coefficient matrix is uncertainty when the number of sample is not symmetry. The recognition algorithm based on PSVM is the first through principal component analysis (PCA) for dimensionality reduction and then use PSVM to classify. In this paper, we make a series of experiment in ORL face database and Yale face database, and analyze the different recognition rate selecting different correction coefficient matrix. The experiment result show that selecting correction coefficient matrix have intimate relationship with abundant degree of facial expression.