Kernel Fisher Discriminant Analysis for Palmprint Recognition

In this paper, a method for palmprint recognition, kernel Fisher discriminant analysis (KFDA), is proposed. The method introduces KFDA to represent palmprint features for palmprint recognition. In the paper, a device without fixed peg is developed to capture palmprint images. Because the movement, the rotation and the stretching of hands are uncontrollable, the features extracted from these palmprint images have a little nonlinearity. Classic linear feature extraction approaches, such as PCA and FLDA, only take the 2-order statistics among palmprint image pixels into account, and are not sensitive to higher order statistics of data. Therefore, KFDA is used to extract higher order relations among palmprint images for future recognition. The experiment results denote that KFDA have a better performance than eigenpalms and fisherpalms, especially in case of using a small quantity of training samples.