基于自适应阈值小波变换的心音去噪方法

A Heart Sound Denoising Method Based on Adaptive Threshold Wavelet Transform

在采集心音信号时,难免会引入一些噪声,对心音信号诊断之前必须对其做去噪处理。由于心音信号是非线性非平稳信号,对心音信号去噪处理常用小波变换去噪方法,但是传统的小波阈值函数去噪方法需要自定义阈值,去噪效果也不理想,且可能会滤除了大量的细节特征,从而无法对心音信号做出正确的判断。为了克服传统小波阈值函数对心音信号去噪处理出现失真的问题,本文在半软阈值函数的基础上提出了基于蚁群算法优化选取阈值的非线性小波变换去噪方法。以原始心音为研究对象,通过选用db6小波并进行6层小波分解,分别选用硬阈值函数、软阈值函数、半软阈值函数、蚁群算法的优化阈值的半软函数等不同的小波去噪处理,并将去噪效果与原始心音进行对比,然后利用蚁群算法的全局搜索性搜索最小均方误差意义下的最佳阈值。仿真结果分析表明:蚁群算法优化选取阈值的心音去噪效果不仅能够去除噪声,还能保留信号细节特征,该方法与传统的硬阈值函数去噪方法相比信号的信噪比(SNR)和均方根误差(RMSE)均得到明显的改善。

In the acquisition of heart sound signal,it is inevitable to introduce some noise so that the heart sound signal denoising must be done before the diagnosis of heart sound signals.Because the heart sound signal is nonlinear and non-stationary,wavelet transform denoising method is commonly used to remove noise of heart sound signal.However,traditional wavelet threshold function needs to customize the threshold,its denoising effect is not ideal,and may filter out a lot of useful details of the heart sound signal,which may hardly lead to a correct judgment.In order to solve the problem of the distortion of the heart sound signal denoising process in using traditional wavelet threshold function,on the basis of semi soft threshold function,a nonlinear wavelet transform denoising method based on ant colony optimization algorithm is proposed.Using the original heart sounds as the research object,by using the DB6 wavelet and 6layer wavelet decomposition,this paper use different denoising methods such as hard threshold function,soft threshold function,semi soft threshold function and ant colony algorithm of the optimal threshold of semi soft function of wavelet denoising,compare the effects of these methods and then use ant colony algorithm global search to search for the optimal threshold in terms of minimum mean square error.Simulation results show that the ant colony optimization algorithm to select the threshold of the heart sound denoising can not only remove the noise,but also preserve the details of signal characteristics,and the method is more effective in noise reduction in comparison with the conventional soft and hard threshold functions.The method is more effective in noise reduction in comparison with the conventional soft and hard threshold function.