Statistical analysis-based DCT method for face recognition

It´s critical for feature recognition of DCT (Discrete Cosine Transform) coefficient selection in the new Data Field which commutated from the DCT; however, the DCT commutating does not have the capability for data compression. This paper proposes a new method combined with the DCT commutates and the statistical analysis to select DCT´s coefficient. Firstly, calculate each capacity of DCT´s coefficients. Secondly, select the data as a feature which has a higher discriminate ability, and then combined with the PCA method for data decorrelation. Finally, do the classification and recognition. Experiments show that this method of classification efficiency greatly improved compared to traditional methods.