Ica And Wavelet Neural Network-based Face Recognition Research

Some questions about the automatic recognition of human face are researched on the thesis. They contain the methods of face detection and face recognition that have been researched. Furthermore, I research and improve the method of face feature abstraction and face recognition, and present the method of face recognition based on ICA and wavelet neural network, the main content of the dissertation is summarized below:We have researched how to abstract face feature by PCA and ICA,contrasted which feature is more better to represent face.The speed of training ICA face base is accelerated by FASTICA method, at the same time, we select better face base using some methods,thus can use as few and better base as possible to represent face,and reduce the dimension of face feature space in effect, afterward,we train wavelet neural network by face feature abstracted using the better ICA face base.when testing,we input face feature abstracted using the better ICA face base for test to the wavelet neural network that has beem trained,and the recognition rate can be obtained by the output. For face feature abstractingjndependent Component Analysis method is more efficient than Principal Component Anaalysis in making use of high rank statistic information of face. Moreover,the wavelet neural network is better than other forward neural network because it can avoid to design network blindly. Otherwise,we improve on how to adjust learning rate of wavelet neural network and set initialization,make it better easy to convergence.Because of the speed of network convergence accelerated,the rate of face recognition, and preciseness is improved on.The result of experiments on ORL face database prove the method of face recognition based on ICA and wavelet neural network is better.