Further research on principal component analysis method of face recognition

A problem which must be considered is the number of eigenvalues reserved, when PCA (principal component analysis) is used to deal with dimension reduction. In this paper, two eigenvalue extraction methods based on PCA algorithm for face recognition are proposed by using Kaiser criterion and eigenvalue curve. The Kaiser criterion method is to discard eigenvalues which are less than one. The eigenvalue curve method is to decide the number of eigenvalues reserved according to eigenvalue curve. Through the experiments on ORL face database, both are compared with the normal threshold method, respectively. The running time of procedure is reduced in the Kaiser criterion method, but the correct face recognition rate isn´t changed. The correct face recognition rate is raised in the eigenvalue curve method, but at the same time the running time of procedure is increased.