Abstract: For wavelets network with high-dimensional inputs,both the initial parameters and network structure are very complicated,and the amount of computation is large.Toward this problem,a new method is proposed using support vector machine(SVM) to ascertain both the initial parameters and network structure of wavelet network.At first,the supervised fuzzy clustering algorithm is used to extract fuzzy rule from clustering.And then,for the consequent rule,SVM is applied to both the initial parameters and the network structure of wavelet network.Finally,gradient descent algorithm is adopted to adjust the parameters of fuzzy wavelet network so as to make the error between the outputs of fuzzy wavelet network and the expectant outputs is less.Simulation result shows that this algorithm can greatly improve the classification precision comparing to traditional fuzzy neural network(FNN).