Face recognition using radial basis function (RBF) neural networks

This paper presents some new results on face recognition using Radial Basis Function (RBF) Neural Networks. First, face features are extracted by discriminant eigenfeatures. Then, a general approach, which determines the initial structure and parameters of RBF neural networks, is presented. A hybrid learning algorithm is used to dramatically decrease the dimension of the search space in the gradient method, which is crucial on optimization of high-dimension problem. Experimental results conducted on the ORL database image of Cambridge University show that the error rate is 1.5% which is a tremendous improvement over the best existing result of 3.83%