Abstract: 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 6 layer 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.

Key words: heart sounds denoising , wavelet transform, adaptive threshold, ant colony algorithm