A novel way of face recogniton to improve the quality of features extraction

This paper sets to improve the representative information of the eigenvector of the face in which Haar wavelet is applied to decompose the face. The high-frequency of the face decomposed on 1 scale is preserved as a high-frequency feature-vector (HF), and the low-frequency of the face decomposition is applied for dimension reduced by Fisher´s linear discriminant(FLD). Then we combined the high-frequency feature-vector and the eigenvector extracted by FLD as a stand-by recognition face eigenvector(SRFE). The data of the SRFE is used to train and to test a fuzzy neural network which is applied for face recognition. A simulation of the proposed algorithm is done on the basis of Olivetti Research Lab(ORL) face database, and the results show that the algorithm is able to recognize quickly with high recognition rate.