Efficient Algorithms for Palmprint Identification Based on Non-linear Feature Extraction and Contourlet Transform
One of robust biometrics is palmprint. Feature extraction from palm area is an important issue and can determine the complexity and
efficiency of identification system. In this paper, we propose a new method to extract region of interest (ROI) and two efficient algorithmsfor
feature extraction from palmprint. The first algorithm calculates isometric projection (IsoP) of ROI and the second one calculates the graylevel co-occurrence matrix (GLCM) of contourlet sub-bands of ROI to extract raw features. In both algorithms linear discriminant analysis (LDA) is used to select proper features from raw extracted features and then,K-nearest neighbor (KNN) and support vector machine (SVM) classifiers are used to identify person. Hong Kong Polytechnic University (PolyU) palmprint database is used to evaluate the performance of the proposed algorithms. Experimental results on 200 different persons demonstrate that the proposed methods have better efficiency in comparison with recently proposed algorithms for palmprint identification.