A Novel Features Extracting Technique Useing Location in Face Recognition

In PCA for the face recognition technique, computing the eigenvalues of the image matrix is the first step. Then these eigenvalues are sorted in descending order. A certain number of vectors, which correspond with these eigenvalues, are chosen by descending order and regarded as features of the face. Furthermore, these features show the whole image's characters. There are differences between the image's features and the face's. In this paper, a novel method based on face facial features specific positioning is proposed. This method consists of four steps: i) Locating the eyes and the middle line of the face, ii) Finding the areas of five sense organs by the face's symmetry, iii) Dimension reduction to each selected area using PCA, iv) Creating the transformation matrix for SVM classification. The recognition rate of this method is higher than the old method's. This suggests that these features from new method are more representative the real face features than old method.