Multi-Layer Perceptron Neural Network Based Person Identification Using Appearance Based Facial Feature

Abstract

This paper deals with secure person identification system using Back-Propagation learning neural network algorithm where appearance based facial feature and Principal Component Analysis based dimensionality reduction technique have been used. To extract the appearance based facial features, Viola-Jones method has used as a face detector, Stam’s method of Active Shape Model has applied for detecting the facial edges and image pre-processing method has used for eliminating the background noises. For reducing the effect of uneven lighting from the facial images, contrast based edge detector has been used for this work. Principal Component Analysis based reduced facial feature vector are feed to Back-Propagation learning neural network algorithm to create the learned template which is finally used to classify the unknown facial pattern. To measure the performance of the proposed system, extended Yale Face Database has been used which focuses the superiority of the proposed appearance based facial feature for person identification system.
Keywords: person identification, appearance based facial feature, principal component analysis, back-propagation learning neural network.