基于经验模式分解的去噪方法

摘要

If using EMD to decompose white noise, the product of energy density and average cycle of the intrinsic mode function component will be a constant, based on this conclusion, the EMD filtering denoising method is developed. The analysis for modeling data shows that EMD can be used as a denoising filter, and the denoising effects for the filter is related to the noise level. When noise's variance is less than signal amplitude and there is no high frequency signal, the filter works very well and the results for filtering and denoising are excellent. The denoising effect for the EMD depends on if there are high frequency signal in the data-sets, if there are high frequency signal and the noise level is high, the obvious distortion in the filtering curve can be found.%根据经验模式分解(Empirical Mode Decomposition,简称EMD)方法分解白噪声而得到的本征模式函数分量的能量密度与其平均周期的乘积为一常量这一特性,本文建立了一种滤波去噪方法,即EMD滤波去噪法.通过模拟数据试验分析表明:EMD可以作为一种去噪滤波器,EMD方法的去噪能力与噪声水平有关.对于噪声方差小于信号振幅且无高频信号时,其滤波去噪的效果良好;EMD方法的去噪能力还与待滤波数据中是否含有高频信号有关,而当噪声水平较大且待滤波的序列中又具有高频信号时,滤波曲线会出现明显的失真现象.