Face recognition using hybrid classifier systems

This paper considers hybrid classification architectures and shows their feasibility on large databases consisting of facial images. Our architecture, consists of an ensemble of connectionist networks-radial basis functions (RBF)-and decision trees (DT). This architecture enjoys robustness via (i) consensus provided by ensembles of RBF networks, and (ii) categorical classification using decision trees. The results reported in this paper on automatic face recognition using the FERET database are encouraging when one considers that the size of our test bed is in excess of 350 subjects and the great variability of the database. In addition we have also demonstrated the feasibility of our approach on queries aimed at the retrieval of frames ('images') using contextual cues.