Genetic and evolutionary methods for biometric feature reduction

In this paper, we investigate the use of Genetic and Evolutionary Computations (GECs) for feature selection, weighting, and hybrid weighting/selection in an attempt to increase recognitionaccuracy and reduce the number of features needed for biometric recognition. The GECs were first applied to a subset of subjects taken from the Facial Recognition Grand Challenge dataset. The resulting feature masks were then tested on a larger subset in an effort to investigate how well they generalise to unseen subjects. The results suggest that our GECs effectively increase the recognition accuracy, reduce the features needed, and create feature masks that generalise well. Copyright © 2012 Inderscience Enterprises Ltd.