“…We respectively random select 3, 4, and 5, samples in ORL face database as training set, the remaining as test set. Our algorithm is compared with the principal component analysis (PCA) [8], linear decision analysis (LDA) [9], independent component analysis (ICA) [10], M-PCNN [15] and the algorithm in literature [17]. In order to eliminate the influence of different classifier for recognition accuracy, each method is classified by support vector machine with the optimum parameters.…”