Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Face Recognition Based on Radial Basis Function Neural Networks Classifier

Author LiZuo
Tutor JingXiaoYuan
School Harbin Institute of Technology
Course Computer Science and Technology
Keywords Fisher linear discriminant analysis Nuclear methods To identify the common vector Radial basis function neural network classifier
CLC TP391.41
Type Master's thesis
Year 2006
Downloads 462
Quotes 2
Download Dissertation

With the development of computer technology and communication technology, the increasingly high demand for the security of personal information. In this case, a new method of authentication - biometric identification technology - a wide range of applications. The face recognition technology is a biometric identification technology, it is the most consistent with the form of the human mind functioning, it has been a hot area of ​​research. The face recognition technology is applied to the actual environment, however, there are still many difficulties, such as the problem of high-dimensional and small sample size problem. This paper attempts to solve these two problems with the extraction method and radial basis function neural network classifier based on the nonlinear characteristics of the Fisher discriminant function on the basis of detailed analysis of the latest research progress and the main features of the face recognition, in order to improve the face recognition the correct rate. The first study of face recognition based the Fisher criterion nuclear discriminant analysis method. Fisher linear discriminant function method is a classical statistical methods can be applied in feature extraction and classification decisions two aspects. However, this method can not achieve good results in nonlinear problems such as face recognition. Therefore, the nonlinear kernel method is applied to the face recognition technology can identify a common vector method has been improved in the nuclear space, a new Approves identification of common vector method (Kernel Discriminative the Common Vectors KDCV ). Endorsed identify the common vector method to extract the common vector of each class of samples in the nuclear space as optimal discriminant vectors effectively extract facial feature high-dimensional face image reduced to M-1 dimensional vector to reduce the computational burden and improve the recognition accuracy rate. The experimental results show that the approval of this article to identify the common vector has achieved good results in terms of facial feature extraction. The second work is to study the radial basis function neural network (Radial Basis Function Neural Networks, RBF) as a classifier in face recognition. Radial basis function neural network is a simple two-layer feedforward artificial neural network, with its advantages of simple structure, fast training process and good generalization ability can be widely used in many fields, especially in the pattern classification and function approximation aspects. Mainly focus on two aspects of the determination of the number of radial basis function neural network center and the choice of the center position, and constructed a follow Fisher discriminant function criteria clustering method. The application of this classifier to face recognition, has been better than the results of other methods. On the basis of two main work, this paper endorsed identify common vector method and radial basis function neural network classifier proposed based on the approval of the RBF neural network classifier to identify common vectors (Kernel Discriminative the Common Vectors plus on RBF Classifier the Approach, KDRA) method. The AR face database and ORL face database experiment. Experimental results show that, compared with the other linear discriminant method, KDRA method can be better

Related Dissertations
More Dissertations