Palmprint recognition with applying different kernel matrix sizes on Gabor wavelet features

Abstract

This paper presents Gabor based Kernel Principal Component Analysis (KPCA) palmprint recognition method for human identification. The intensity values of palmprint images extracted by using an image preprocessing method are first normalized. Then these images are transformed to the spectral domain by using Gabor wavelet transform. The transformed palm images exhibit strong characteristics of spatial locality, scale, and orientation selectivity. Next, the feature vectors are nonlinearly maps into a high dimensional feature space with KPCA method. In this method during kernel matrix calculation, the sample numbers per class changed and it's effect investigated. Finally, weighted Euclidean distance based nearest neighbor method is realized for classification. The proposed algorithm tested on the most-well known palmprint database, PolyU, includes 7752 samples of 386 different people.