Analysis of Training Functions in a Biometric System

Abstract

One of the commonly used Biometric methods is Face Classifica tion . Face images a re obtained from FEI face database. In this paper, different Training F unctio ns of Neural Network a re studied . In this research, a face recognition sy stem i s suggested based on feedforward backpro pagation neural network Neural Network (BPNN) model. Each model i s constructed separately with one input layer, 3 hidden layers and one output layer). Four ANN training algorithms (TRAINLM, TRAINBFG, TRAINGDX, and TRAINRP ) a re used to train each model se parately. Performances using each of the training algorithms were evaluated based on mean square error an d the best training algorithm i s found for the face recognition.