Face Recognition by Combination of RBF Neural Networks Using Dempster-Shafer Theory

This paper presents an approach to face recognition based on Dempster-Shafer (DS) theory of evidence, which combines the evidences of two radial basis function (RBF) neural networks. The degrees of belief of the two RBF neural networks for classification of an image have been estimated using two different feature vectors derived from images of the ORL face database. Then these degrees of belief have been combined using DS theory to improve the overall recognition rates. The average recognition rates of the proposed method have been found to be 83.78%, 88.08%, 97.10%, 98.06% and 97.75%, in 10 different experimental runs of 3, 4, 5, 6 and 7 training images out of 10 images per individual, respectively. The proposed method is found to be better than some of the existing methods