Face Recognition Algorithm Based On Kernel Discriminant Analysis

Automatic face recognition is a typical pattern analysis, understanding and classification problem, which is closely related to many disciplines such as pattern recognition, image processing, computer vision, statistical learning, and cognitive psychology etc. The in-depth study and final settlement of AFR can greatly promote the maturity and development of this disciplines. As one of the main research areas in Biometrics, face recognition is believed having a great deal of potential applications in security system, human ID, digital surveillance and so on. Researching on the face recognition technology has great theoretical and practical values.The development and mainly methods of face recognition technique are first introduced in this paper. Then the problem of preprocessing, feature extraction and classification are discussed.The aim of face image preprocessing is to regularize the face image which is captured by image collecting devices to normalized image, it includes two steps mainly: geometry normalization and grey value normalization.A Central issue to a successful approach for face recognition is how to extract discriminate feature from the facial images. Linear and nonlinear kernel projection algorithms applied for feature extraction are focus on. As for linear projection, Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) are expounded. As far as Small Sample Size problem, this paper proposed a new method based on PCA and LDA.In many real-world applications the distributions of original samples are usually highly complex and nonlinear. We take advantage of the kernel method in Support Vector Machine (SVM) and research on Kernel Principal Component Analysis (KPCA) and Kernel Fisher Discriminant Analysis (KFDA). However, in KFDA the total scatter matrix is required to be full rank, which is seldom satisfied in the face recognition tasks. In order to overcome this problem, the orthogonal complementary space method based on kernel method is presented in this part. In the new approach, the kernel method is used firstly to project the original samples into an implicit space called feature space by nonlinear kernel mapping, then two equivalent models based on generalized Fisher criterion have established by the theory of reproducing kernel in the feature space, and the optimal discriminant vectors are solved finally by using the technique of orthogonal complementary space.After the features of the images are extracted, a back propagation (BP) algorithm is introduced to train the neural network for recognition. To verify the effectiveness of the proposed method, experiment is tested on ORL face database. The experiment results demonstrate the effectiveness of the proposed methods.