Abstract: Hard threshold is poor in denoising, and soft threshold makes signal distortion because of smooth transition, so a new adaptive-threshold denoising method is presented. The algorithm by using soft threshold with the improved threshold to denoising the signal combines the mathematical statistics with the distribution of signal and noise wavelet coefficients and transmission characteristics and the contribution rate of noise standard deviation. The results show that the improved algorithm makes better MSE performance and SNR gains than conventional threshold algorithm and some improved threshold algorithm. And a good performance at remaining the characteristics information of power quality signals.

Key words: wavelet transform, threshold denoising, power quality, Signal Noise Ratio（SNR）

摘要： 针对硬阈值去噪效果不佳，软阈值过度光滑使信号失真的缺点，提出了一种改进的自适应的阈值去噪方法。该算法将数理统计与信号和噪声的小波系数的分布规律和传播特点及噪声的标准差贡献率结合起来，对阈值门限进行了改进，并采用软阈值函数对信号进行处理，实现了其去噪的功能。结果表明，改进后的算法对电能质量信号进行消噪处理，信噪比增益和均方误差上均优于传统阈值算法及一些改进后的阈值算法，而且能够较好地保留电能质量信号的特征信息。

关键词: 小波变换, 阈值去噪, 电能质量, 信噪比