A Hybrid Face Detector Based on an Asymmetrical Adaboost Cascade Detector and a Wavelet-Bayesian-Detector

Biologically inspired receptive fields arc used to process input facial expressions in a modular network architecture. Local receptive fields constructed with a modified Hcbbian rule (CBA) arc used to reduce the dimensionality of input images while preserve some topological structure. In a second stage, specialized modules trained with backpropagation classify the data into the different expression categories. Thus, the neural net architecture includes 4 layers of neurons, that we train and test with images from the Yale Faces Database. A generalization rate of 82.9% on unseen faces is obtained and the results are compared to values obtained with a PCA learning rule at the initial stage.