经验模式分解端点效应抑制方法的研究

Research on Method for End Effects Reduction of Empirical Mode Decomposition

针对经验模式分解法(Empirical Mode Decomposition,EMD)中的端点效应严重影响算法精度的情况,为了减小端点效应,文中提出一种新的先匹配后镜像延拓方法。该方法借鉴镜像对称延拓与波形特征匹配延拓,将信号先进行波形特征匹配延拓再进行镜像对称延拓。利用仿真数据对先匹配后镜像延拓的方法进行了验证,并与原有几种延拓方法在不同评价指标下进行了对比。结果表明,先匹配再镜像延拓后,信号的包络线发生畸变最小,同时新方法分解得到的IMF精度较高,正交性好。先匹配后镜像延拓方法提高了经验模式分解的精度,能更有效地抑制经验模式分解法中的端点效应。

The precision of Empirical Mode Decomposition （EMD） is reduced greatly by its end effects, so a new end extending method combining the waveform feature matching extending method and mirror extending method is presented. The waveform feature matching extending method and mirror extending are used for reference in this new method,the signal is extended by waveform feature matching method at fast and then extended by mirror method. A simulation signal is applied to test the performance of the new method, and a comparison under different evaluating indicators between the new method and old methods is made and analyzed. The result shows that the signal envelope has minimum distortion, at the same time, the IMF has high precision and good orthogonality decomposed by new method. The proposed improved extending method can improve the precision of EMD and restrain the end effect effectively.