消除经验模态分解中混叠现象的改进掩膜信号法

Use of masking signal to improve empirical mode decomposition

经验模态分解方法能在时频域上正确地描述非平稳非线性信号的局部特征。但由于模态混淆,当信号组合分量的频率太接近时,常不能得到正确的经验模态分解结果。针对这一情况,提出了消除经验模态分解中混叠现象的一种方法——改进的掩膜信号法,并将其应用于风机叶片振动信号的分析中。该方法以能量为基础对掩膜信号的选择进行改进,并通过掩膜信号结合EMD来达到消除模态混叠现象的效果。对风机叶片振动信号进行验证的结果表明,该方法简便易行,可有效分离混叠模态,提取有用信号,并且对白噪声也有削减效果。

The empirical mode decomposition(EMD) was applied to deal with non-stationary signals,and to describe the time-frequency characteristics.However,EMD yields its own interpretation of combinations of pure tones,and mode mixing is one of the perplexing problems.A solution involving an improved masking signal was proposed.The improved masking signal method changes the way to choose the masking signal,and the IMF1 without mode mixing was then achived.The method also allows EMD to be used to separate similar components in frequency domain that couldn't be seperated with standard EMD techniques.The experiment shows that the approach is practical and effective.