Performance Analysis of Fusion Classifier of Face Recognition System based on ANN and GA

Abstract

Though a lot of significant work has been done for recognizing and improving the performance in facial pattern, most techniques have been developed based on a single classifier. But most often, the performance of the classifier is not satisfactory in some case. It has been proposed that the performance of face recognition will beimproved by fusing two or more classifiers. We can combine the information/data in different level i.e., feature levels, data levels or classifier levels. We developed two classifier named back-propagation neural network and genetic algorithm. Then we combined their output based on newly developed fusion algorithm. The performance ofthe proposed fusion classifier can give better recognition rate compared with the performances of individual classifier. Here the fusion has been occurred in two stages. First we have selected the close classified facial pattern for testing pattern by using backpropagation classifier and genetic algorithm classifiers. Then final facial image has been selected among the previously selected facial image by calculation some feature and distance from the test facial pattern.Keywords: fusion classifier; back-propagation algorithm; genetic algorithm;principal component analysis