Two dimension locality preserving projections with class information for face recognition

The dimension reduction is necessary steps for face recognition based on subspace analysis. The proposed method employs class information for structure of similarity matrix when implement of 2DLPP. A subspace which preserves local neighbor structure and centralizes same class samples of training images is got. Moreover, it has available computation efficiency and accuracy because it belongs to the methods based on images which avoid the matrix singularity problem. The performance of the proposed method is evaluated and compared with other popular subspace analysis method based on ORL database. The experiment results show that it has more accurate recognition than previous methods.