Palmprint recognition using principle component analysis implemented on TMS320C6713 DSP processor

Abstract

This paper presents a human identification system using eigen-palm images. The proposed method consists of three main stages. The preprocessing stage computes the palmprint images to capture important information and produce a better representation of palmprint image data. The second stage extracts significant features from palmprint images and reduces the dimension of the palmprint image data by applying the principal component analysis (PCA) technique. Low-dimensional features in the feature space are assumed to be Gaussian. Thus, the Euclidean distance classifier can be used in the matching process to compare test image with the template. The proposed method is tested using a benchmark PolyU dataset. Experimental results show that the best achieved recognition rate is 97.5% when the palmprint image is represented using 34 PCA coefficients. Moreover, the Euclidean distance classifier is implemented on a digital signal processor (DSP) board. Implementing the proposed algorithm using the DSP processor achieves better performance in computation time compared with a personal computer-based system