中文摘要：

多小波可以同时满足对称性,短（紧）支撑性,二阶消失矩和正交性,在信号处理方面比小波更有优势。在多小波分解系数中,当一个系数包含某些信号成分时,其相邻系数也可能包含一定的该信号成分。与传统的小波去噪算法相比,相邻系数法能更好地解决变换系数之间的相关性,获得更好的降噪效果。在模拟振动信号降噪处理中,对信噪比和均方根误差进行定量分析,结果表明多小波相邻系数降噪法要优于小波阈值降噪法。在对水电机组实际采集的振动信号进行降噪研究时,通过对比降噪前后信号振动特征分量的保持完整程度,表明该方法滤波更彻底,保留振动特征分量更完整,是一种更加有效的降噪方法。

英文摘要：

Multiwavelets have such features as short support,orthogonality,symmetry and vanishing moments.Therefore,multiwavelets offer the possibility of superior performance for signal de-noising,compared with wavelets.After multiwavelets decomposition,if the current coefficient contains certain components of the signal,its neighboring coefficients may also contain these ingredients.Based on this,we put forward the multiwavelets neighboring coefficients de-noising method.We apply this method to hydropower unit vibration signal de-noising,settling down the contradiction between restraining noisy pollution and reserving signal details by the traditional wavelets threshold de-noising.After de-noising the simulation signal,we find that multiwavelets neighboring coefficients de-noising method gives better performance both at SNRand RMSEthan wavelets threshold de-noising method.For the factual hydropower unit vibration signal,according to the integrity degree of vibration characteristic components,we discover that multiwavelets neighboring coefficients de-noising method can filter the noise more thoroughly and retain the characteristic ingredients more completely,showing a better performance.