Research On Method Of Human Face Recognition

As human face recognition has the characteristics of being more direct, safer, more effective, so that it becomes an identity recognition technique that is easily accepted and has important research and application values. Meanwhile, it has been an active research topic in the pattern recognition filed.In this paper, the preprocessing method for face recognition is presented in detail firstly, which includes the rough location and precise location of human eyes, slant correction, human region cropping, size normalization, gray compensation, etc. For precise location of the eyes, a method that combines geometric characteristics with Sobel operator and Hough transformation is used to locate human eyes and precise location of the eyes is obtained. Bilinear interpolation is used to normalize the size, and histogram equalization is used to obtain gray compensation.Modular principle component analysis (MPCA) algorithm and singular value decomposition (SVD) algorithm for face recognition are discussed in detail. MPCA can extract features which have a strong ability to identify human. However, as a sub-space method, MPCA is sensitive to translation, rotation and other geometric transform so that its recognition rate gets lower when changes in light intensity and facial expression is radical. SVD algorithm is an effective algebraic feature extraction method. The singular value feature has the advantage of translation invariance and rotation invariance. But SVD has not information enough to identify faces. Therefore, a human face recognition method of combing MPCA with SVD is proposed to expect higher recognition rate. The geometry invariance of SVD can compensate the geometry sensitivity of MPCA, and the advantage of MPCA strong ability to identify faces can make up the disadvantage of SVD inadequate information to identify faces. Experimental results show that the proposed method can obtain higher recognition rate.