Fusion of multiple texture representations for palmprint recognition using neural networks

Abstract

During the last decade, palmprint recognition has received an increasing attention due to the abundant features that can be extracted from the captured palmprint image. However, a single palmprint texture representation may not be sufficient for reliable recognition. Therefore, in this paper we propose a computational model for palmprint recognition/identification by fusing different categories of feature-level representations using a multilayer perceptron (MLP) neural network. Features are extracted using Gabor filters and Principle Component Analysis (PCA) is used to reduce the dimensionality of the feature space by selecting the most relevant features for recognition. The proposed model has shown promising results in comparison with naïve Bayes, rule based and K*algorithm. © 2012 Springer-Verlag.