Design of face recognition algorithm using hybrid data preprocessing and polynomial-based RBF neural networks

Abstract

This study introduces a design of face recognition algorithm based on hybrid data preprocessing and polynomial-based RBF neural network. The overall face recognition system consists of two parts such as the preprocessing part and recognition part. The proposed polynomial-based radial basis function neural networks is used as an the recognition part of overall face recognition system, while a hybrid algorithm developed by a combination of PCA and LDA is exploited to data preprocessing. The essential design parameters (including learning rate, momentum, fuzzification coefficient and feature selection) are optimized by means of the differential evolution (DE). A well-known dataset AT&T database is used to evaluate the performance of the proposed face recognition algorithm. © 2012 Springer-Verlag.