一种改进的集合平均经验模态分解去噪方法

针对现有非平稳信号去噪处理中的难点，提出一种改进的基于集合平均经验模态分解（EEMD）去噪方法，该方法根据本征模态函数（IMF）能量从被测信号中估计出噪声，使用估计的噪声代替EEMD方法中添加的噪声，最后将集合平均IMF分量累加得到去噪信号。使用相关性判据剔除了EMD分解产生的伪IMF分量，改进了噪声估计方法。仿真表明，改进的方法能够对调幅调频含噪信号进行有效的去噪处理。

An Improved EEMD-based Denoising Method

An improved EEMD-based denoising method is proposed for denoising processing of non-stationary signals. In this method, the noise is estimated and extracted from the measured signals based on the intrinsic modal function （IMF） energy method, and used to replace the additional noise in the EEMD method. Then, all the accumulative mean IMF components and the residuals are summed up to get the denoised signals. A correlation criterion is used to eliminate the pseudo IMF components after each EMD decomposing, and the noise estimation procedure is improved. Simulation results prove that the improved method is effective in denoising of AM-FM noise signals.