带噪语音信号小波去噪算法研究

去噪问题是信号处理中必不可少的问题。滤波作为传统的去噪方法，主要包括线性滤波和非线性滤波。噪声与信号频率重叠，传统方法要想取得良好的去噪效果必须牺牲部分信号。现有的小波模极大值去噪方法虽然有较好的去噪效果，但是该方法计算量大。小波阈值去噪方法更简单，但小波重构后的小波系数与噪声的小波系数存在恒定的偏差。在阈值去噪方法的基础上提出一种改进算法，仿真结果表明该算法在加性高斯白噪声污染下表现出较好去噪效果。

Research on wavelet-based speech signal with noise denoising algorithms

Denoising is an essential problem in signal processing. Filter method as the traditional denoising method, includes linear and nonlinear filtering. Due to?frequency ?overlap between noise and signal, the traditional method to obtain good denoising effect must sacrifice some signal. Existing wavelet modulus maxima denoising method has better denoising effect, however?the computation loads are high. Wavelet threshold denoising method is simpler, but the wavelet coefficient after wavelet reconstruction has biased estimation with the wavelet coefficient of noise. The paper analyzes the threshold denoising method, giving a method to improve, simulation results show that it has better denoising effect.