Palmprint Recognition by using Bandlet, Ridgelet, Wavelet and Neural Network

Abstract

Palmprint recognition has emerged as a substantial biometric based personal identification. Tow types of biometrics palmprint feature. high resolution feature that includes: minutia points, ridges and singular points that could be extracted for forensic applications. Moreover, low resolution feature such as wrinkles and principal lines which could be extracted for commercial applications. This paper uses 700nm spectral band PolyU hyperspectral palmprint database. Multiscale image transform: bandlet, ridgelet and 2D discrete wavelet have been applied to extract feature. The size of features are reduced by using principle component analysis and linear discriminate analysis. Feed-forward Back-propagation neural network is used as a classifier. The recognition rate accuracy shows that bandlet transform outperforms others.