Please use this identifier to cite or link to this item:
|Face Recognition based on Wavelet Transform, Principal Component Analysis and Neural Network
|Nigam, Rahul Dev
|Singla, Sunil Kumar
|Face Recognition;Neural Network;Wavelet Transform, Principal Component Analysis
|Face is a primary focus of attention in social intercourse, playing a major role in conveying identity and emotion. The dictionary meaning of FACE is “The surface of the front of the head from the top of the forehead to the base of the chin and from ear to ear”. Face recognition records the spatial geometry of distinguishing features of the face. Face recognition has been an interesting issue for both neuroscientists and computer engineers dealing with artificial intelligence (AI). A human can identify face easily, whereas, for a computer to recognize the face, the face area should be detected, features are required to be extracted and comparison with the features already stored in the database is required. A key potential advantage of a machine system is its memory capability, whereas, human face recognition system the important feature is the parallel processing capability. In this work, new face recognition system based on wavelet transform and principal component analysis using back propagation algorithm neural network has been presented. The HAAR is used to form the coefficients matrix for the detection of the face. The image feature vector is obtained by computing principal component analysis from the coefficient matrix of discrete wavelet transform. The Eigen faces approach is then used to reduce the dimension of the face vectors. Reduced feature vector are used for further classification using neural network. Neural network is used to create the face database and recognize and authenticate the face by using the weights. The proposed work deals with two problems, namely facial expression and illumination. HAAR wavelet transform enhances the contrast as well as edges of the face images and works efficiently in wide range of illumination changes. Our experiments have been conducted on the YALE database to obtain the optimum learning rate which comes out to be 0.5 in this case considering the final goal and the success rate. The recognition rate obtained is 97.92% when 52 components from PCA have been selected and 50 neurons in the hidden layer have been used.
|M.E. (Electronic Instrumentation and Control)
|Appears in Collections:
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.