Identification Information Analysis of Sample Train Set Subspace

Principal component analysis (PCA) which is widely used in pattern recognition field aims at reducing the dimension of sample. PCA replaces variables in the original sample vectors that have redundant information with fewer integrative variables. The recognition ability used author´s algorithm is tested in the paper. It is proved that zerospace do not include any identification information which would be useful for distinguishing different samples. Experiment results based of our lab´s facebase and ORL face base shows the theory is right.