改进的EMD算法及其在含噪盲扰信分离中的应用

针对经验模态分解(EMD)降噪算法存在因表示噪声的内蕴模态函数(IMF)分量选择不当而引起的降噪性能不稳定的问题,对其进行改进,加入对噪声水平的估计,并将噪声阈值作为分解结束的判决门限,避免了对噪声IMF分量的选择。在此基础上,联合小波变换进行降噪,并应用于含噪盲扰信分离中。仿真表明在一定范围的低信噪比条件下,该算法增强了盲扰信分离的抗噪声性能,可将源信号从染噪的观测信号中有效地分离出来。

Improved EMD Algorithm and Its Application to Blind Interference-signal Separation with Noisy

We propose an improvement to the empirical mode decomposition （EMD） method, which needn＇ t choose noise components and eliminate the unstability of denoise owing to improper noise components by adding an estimation of the noise level and setting a threshold of noise as the end of the decision threshold decomposition. On this base, combining with the wavelet transform to reduce noise and using in the blind interference - signal separation （BISS） for noisy mixtures. Simulation results show that,in a certain range of low signal noise rate （SNR）, this algorithm boosts the BISS ＇s capability to resist noise and can recover the sources from noisy mixture signals effectively.