Face Recognition Using Kernel Fisher Linear Discriminant Analysis and RBF Neural Network

A new face recognition method is presented based on Kernel Fisher's Linear Discriminant Analysis (KFLDA) and Radial Basis Function Neural Network (RBFNN). First, the principal component analysis (PCA) technique is used to reduce the dimension of the facial image. Next, the reduced images are further processed by the KFLDA. Here, KFLDA is used for extraction of most discriminating features in appearance-based face recognition. KFLDA provides better generalizations taking higher order correlations into account rather than FLDA, which projects directions, based on second order statistics. RBFNN is used as a classifier, which classify the face images based on these extracted features. We have tested the potential of the proposed method on the ORL face database. The experimental results show that the proposed method provides higher recognition rates in comparison to some other existing methods.