论文部分内容阅读
小波神经网络是一种新的信号逼近和分类工具．对小波基函数的选取、参数初值确定、参数的计算方法以及在信号逼近和分类中的应用等问题已经有大量研究．小波神经网络应用于信号分类仍有不少具体问题．本文针对模板和样本信号周期不同时网络参数差异较大的问题，对网络结构进行了修正，提出了自适应周期小波神经网络（APWNN，AdaptivePeriodWaveletneuralNetwork），给出了计算方法．以特定人元音识别试验为例，给出了APWNN在信号分类中的具体应用．
Wavelet neural network is a new signal approximation and classification tool. A large number of studies have been conducted on the selection of wavelet basis functions, the determination of initial parameters, the calculation methods of parameters and the application in signal approximation and classification. Wavelet neural network applied to signal classification is still a lot of specific problems. In this paper, aiming at the problem that the network parameters are greatly different when the periods of the template and the sample signal are different, the structure of the network is modified. An Adaptive Periodic Wavelet Neural Network (APWNN, AdaptivePeriodWaveletneuralNetwork) is proposed and the calculation method is given. Taking a particular vowel recognition test as an example, the application of APWNN in signal classification is given.