Face recognition based on MPCA
Caiming Chen, Shiqing Zhang, Yuefen Chen · 2010
In this paper, A new method of face recognition based on multilinear principal component analysis (MPCA) is proposed. First, instead of transforming matrices into vectors for principal component analysis (PCA), the MPCA can use matrices or higher-order tensors directly to capture most variance for dimensionality reduction. The total scatter can be maximized by optimizing the projection matrix. Then all entries from the resulted matrix are sorted according to their class discrimination power for feature selection. Last, the nearest neighbor classifier is used to recognize different faces from the ORL face database. The best accuracy rate can reach 97%.