Algorithms selection and adaptation in accord with architecture for RBF neural network based face authentication SoC

This paper describes the algorithms applied to a Radial Basis Function (RBF) neural network. This neural network is used as a classifier to design a human face authentication system. The aim of this project is to obtain a low cost system on chip (SoC) to replace password identification for one person on mobile devices. Several parts of the neural network need to be modified to obtain good performances for this application with cost limited hardware resources. For the system design, the algorithms are selected and adapted in accord with architecture implantation on hardware platform (methodology AAA). Also we present different adaptations made for a full integration of RBF neural network (learning and recognition steps).