基于改进EMD的信号降噪方法

经验模态分解(Empirical Mode Decomposition,EMD)算法作为新型时频分析方法受到广泛关注,它基于信号的极值特性处理信号,具有自适应强、无需预先确定基函数的优点。但EMD算法本身仍存在模态混叠及EMD强制降噪法易导致信号失真等一系列问题。针对EMD算法的缺陷,提出基于自相关函数的集合经验模态分解方法(Ensemble Empirical Mode Decomposition,EEMD)与小波阈值降噪相结合的改进算法。首先利用自相关函数对高频固有模态函数(Intrinsic Mode Function,IMF)进行选择,然后利用小波阈值降噪法为EEMD设定阈值,最后将改进算法用于信号降噪,并与快速傅里叶变换(FFT)算法、小波阈值算法以及EMD强制降噪算法进行比较。该方法的优点是克服了EMD算法的不足,避免了模态混叠现象,有效地保留了高频信号中分量,降噪效果更好。

Signal Denoising Method Based On Improved EMD

Empirical mode decomposition algorithm is widely concerned with the new time frequency analysis method.It is based on the characteristic of signal processing,which has the advantages of strong adaptability and no need to determine the basis function in advance.However,there are still a series of problems,such as mode mixing and signal distortion because of EMD noise reduction method.In view of the defects of EMD algorithm,this paper proposes an improved algorithm based on the combination of Ensemble Empirical Mode Decomposition and wavelet threshold denoising algorithm. First of all,it uses the correlation function to choose the high frequency of Intrinsic Mode Function;Then,the wavelet threshold is setting a threshold for the EEMD;Finally,the improved algorithm is used for signal denoising and compared with the Fast Fourier Transform algorithm,wavelet threshold algorithm and forced denoising of EMD.The advantages of the method overcomes the shortcomings of the EMD algorithm and avoids the mode mixing phenomenon.It also effectively retains the high frequency signal component.The noise reduction effect is better than the previous method.