Extending the feature set for automatic face recognition

Automatic face recognition poses an exacting problem in feature extraction, pattern analysis and recognition. Several established systems cater for small face populations and have identified a number of features useful for recognition purposes, but these systems and the features used appear inadequate for large face populations. The paper concentrates on improved feature definition and extraction to lead to a fully structured recognition system based on a single frontal view, and is aimed to provide satisfactory recognition within large populations of human faces. This includes a novel profile description achieved from a frontal view of a face and described by a Walsh power spectrum, use of the face contour which is extracted by iterative curve fitting and described by Fourier descriptors, a set of moments describing eye region statistical features and generic geometrical measurements across the face