Feature Selection of Hand Biometrical Traits Based on Computational Intelligence Techniques

摘要

This chapter presents a novel methodology for using feature selection in hand biometric systems, based on genetic algorithms and mutual information. The aim is to provide a standard features dataset which diminishes the number of features to extract and decreases the complexity of the whole identification process. The experimental results show that it is not always necessary to apply sophisticated and complex classifiers to obtain good accuracy rates. This methodology approach manages to discover the most suitable geometric hand features, among all the extracted data, to perform the classification task. Simple classifiers like K-Nearest Neighbour (kNN) or Linear Discriminant Analysis (LDA) in combination with this strategy, getting even better results than other more complicated approaches.