Jos - Ruiz-Pinales, Juan Jorge Acosta-Reyes, and Ren - Jaime-Rivas

In this paper, we present a feature set selection methodology for the recognition of upright faces by means of support vector machines. We perform face recognition in 3 stages: feature extraction, sub-sampling and classification. We employ a measure based on mutual information in order to select the feature sets. The probabilities required for the computation of the mutual information are estimated by using support vector machines.