Palmprint Recognition Based on Subclass Discriminant Analysis

Abstract

Subspace-based palmprint recognition methods, such as principal components analysis and linear discriminant analysis, assume that each class can be grouped in a single cluster. However, this assumption is not reasonable at some situations where a class is assembled in two or more clusters. In order to solve this problem, a novel palmprint recognition method based on subclass discriminant analysis is proposed in this chapter. Each palmprint class is divided into a set of subclasses that can be separated easily in the new subspace representation. After that, the Euclidean distance and nearest neighbor method are employed as the similarity measurement. Experimental results conducted on a database of 86 hands (10 impressions per hand) show that the equal error rate (EER) of the proposed method yields promising result of EER = 0.67 % for verification rate, which demonstrates that the proposed method is effective to solve the problem mentioned above.