Extraction of Facial Features for Recognition using Neural Networks

The extraction of facial features for classification is sought via constrained neural networks using a hybrid of supervised and unsupervised network. This leads to a low dimensional new representation of the gray level images which allows fast and simple classification. The classification scheme is preceded by preprocessing devoted to reducing the viewpoint and scale variability in the data. 1 Introduction Machine recognition of faces is becoming more and more popular and the need for accurate and robust performance is increasing. One can envision having plastic money identification, electronic keys of various sorts, and security identification, all based or supported by face recognition. For these purposes it is crucial that the number of substitution errors will be minimal, since such errors may be very costly. A successful method has to be insensitive to the visual environment in which faces are presented, thus illumination conditions, background, and location variability should n..