The Application Of Dynamic Fuzzy Neural Networks And Research

Fuzzy neural network is one of the most important subject which develops very quickly in the field of information science. Fuzzy neural networks integrate fuzzy systems knowledge which has a obviously expression and powerful learning function, with a learning function and adaptive ability, and is widely used in industrial process control, machine control, consumer electronic, system identification, pattern recognition, image processing, data mining and other fields. Nowadays the research on fuzzy neural network focused on the expression and parameters, but the neural network identification problem is not resolved.The thesis studies dynamic fuzzy neural network and its applications in-depth research based on previous research. The main points are as follows:(1) An improved facial feature extraction methods which combined with state estimation and tensorface algorithm is proposed for multi-view face recognition. Tensorfaces algorithm is an effective mathematical model which can analyze and express the frames of multi-view face images. At first, extract the feature of human face image by face recognition method based on state estimation and tensorfaces algorithm, then use the dynamic fuzzy neural network classifier to recognize, the results show that the accuracy rate of face recognition method based on state estimation and tensorfaces algorithm is higher than the original tensorface algorithm and PCA algorithm, at the same time the new algorithm lower the time-cost.(2) In the article, an improved dynamic fuzzy neural networks (dynamic compensatory fuzzy neural networks) was proposed which combined with the dynamic fuzzy neural networks and compensatory fuzzy neural. Through the ORL face database and the Weizmann face database of many experiments show that the dynamic compensatory fuzzy neural network classifier is superior to dynamic fuzzy neural network classifiers and radial basis function neural network classifier.(3) Because of the dynamic fuzzy neural network by is not very intelligibility, and the number of generated rules and the deleted rules is easy to overlap , this article also researches the generalized dynamic fuzzy neural networks in-depth. Generalized dynamic fuzzy neural network (GDFNN) make up the original defects in dynamic fuzzy neural network to give users more convenient, and the experiment results in function approximation and system identification also show that GDFNN is better than DFNN, and the learning speed has improved significantly.