Designing multiple classifier systems for face recognition

Abstract

Face recognition systems often use different images of a subject for training and enrollment. Typically, one may use LDA using all the image samples or train a nearest neighbor classifier for each (separate) set of images. The latter can require that information about lighting or expression about each testing point be available. In this paper, we propose usage of different images in a multiple classifier systems setting. Our main goals are to see (1) what is the preferred use of different images? And (2) can the multiple classifiers generalize well enough across different kinds of images in the testing set, thus mitigating the need of the meta-information? We show that an ensemble of classifiers outperforms the single classifier versions without any tuning, and is as good as a single classifier trained on all the images and tuned on the test set. © Springer-Verlag Berlin Heidelberg 2005.