Face recognition using Kernel-based Fisher Discriminant Analysis

Abstract

Fisher Linear Discriminant Analysis (FLDA) has been successfully applied to face recognition, which is based on a linear projection from the image space to a low dimensional space by maximizing the between-class scatter and minimizing the within-class scatter. However, face image data distribution in practice is highly complex because of illumination, facial expression and pose variations. We present the use of Kernel based Fisher Discriminant Analysis for face recognition. The kernel trick is used firstly to project the input data into an implicit space called feature space by nonlinear kernel mapping, then Fisher Linear Discriminant Analysis is adopted to this feature space, thus a nonlinear discriminant can be yielded in the input data. Another similar Kernel-based method is Kernel PCA, in which PCA is used in the feature space. The experiments are performed with the polynomial kernel, and this method is compared with Kernel PCA and FLDA. Extensive experimental results show that the correct recognition rate of this method is higher than that of Kernel PCA and FLDA.