Inner-Knuckle-Print Verification System by the Combination of Gabor Wavelet and Wavelet Statistical Features

the inner knuckle print is one of the reliable and new physiological characteristics among different approaches in biometric system. In this paper, we considered the inner knuckle print as a texture and applied two types of feature extraction methods, namely Gabor wavelet filters and wavelet statistical features. Among all features extracted by these approaches, fifty superior features were selected by the forward feature selection algorithm. Features are classified by new method of reference features in order to achieve higher resolution by using K-nearest neighbor, fuzzy K-nearest neighbor, Parzen window and support vector machine classifiers. Database of Contact-free 3D/2D Hand Images with 1770 hand samples from 177 subjects was selected from Hong Kong Polytechnic University. Equal Error Rate of 6.79%±1.74 and accuracy of 95.21%±1.74 were obtained by k-nearest neighbor classifier.

the inner knuckle print is one of the reliable and new physiological characteristics among different approaches in biometric system. In this paper, we considered the inner knuckle print as a texture and applied two types of feature extraction methods, namely Gabor wavelet filters and wavelet statistical features. Among all features extracted by these approaches, fifty superior features were selected by the forward feature selection algorithm. Features are classified by new method of reference features in order to achieve higher resolution by using K-nearest neighbor, fuzzy K-nearest neighbor, Parzen window and support vector machine classifiers. Database of Contact-free 3D/2D Hand Images with 1770 hand samples from 177 subjects was selected from Hong Kong Polytechnic University. Equal Error Rate of 6.79%±1.74 and accuracy of 95.21%±1.74 were obtained by k-nearest neighbor classifier.