Multilevel Face Recognition System

The problem of biometric person identification on the basis of component-based face recognition is considered. It is shown that the face recognition system can be represented as a hierarchically organized multilevel system, in which an ensemble of local classifiers forms “soft” decisions about the images of individual components of a face belonging to given classes. Then, based on the integration of these decisions, the final decision on whether the recognized face belongs to one of the given classes is formed. The problems of constructing a local classifier model, as well as choosing an integrator of intermediate solutions of local classifiers, are formulated and solved.