小波神经网络阈值自学习在信号去噪中的应用

针对某一确定数据采集系统中小波去噪时的阈值选择,提出以小波神经网络加标准信号来标定去噪阈值的方法,从而提高对信号的去噪性能。对于确定的数据采集系统,信号噪声主要来源于系统本身,而且在短时间内系统可视为时不变的。首先给系统一个标准信号输入,将系统的输出输入到小波神经网络,在给定的噪声熵下训练网络使其熵最小,从而得到相应的去噪阈值,仿真实验表明该方法较一般的去噪方法效果好。

Application of wavelets neural network threshold self-study in signals denoising

This article proposed a method to mark denoising threshold from study function of wavelet nerve network in order to improve performance of denoising to signals.Signals noise mainly comes from system itself,and in a short time the system can be regarded as constant.Give the system a standard signal input first,then input the exportation of the system to a neural network,under a noise entropy that be given train the network to make its entropy minimum,thus the corresponsive denoising threshold can be obtained.Simulation results show that this approach is better than general denoising methods.