Facial Detection based on PCA and Adaptive Resonance Theory 2A Neural Network

The prominent feature of Adaptive Resonance Theory neural network is its ability to cluster arbitrary number of input patterns. In this paper, we investigate the performance of ART-2A acting as a classifier in a face recognition system. The Olivetti-Oracle Research Lab (ORL) database of 400 facial images of 40 subjects is used for training and evaluation, and the performance of this network is compared to the MLP and RBF counterparts. The Principal Components of facial images have been used as inputs for the proposed networks. The experimental results reveal that the ART-2A network offers better recognition accuracy even when the illumination of the images is varied or they are corrupted by additive random noise. Moreover the training procedure of the ART-2A is much less time consuming than the other neural networks and its adaptation is fast while introducing a new sample