基于核主分量分析的经验模式分解及其工程应用

若信号的信噪比较小,经验模式分解不能正确分解出基本模式分量,分量中含有伪分量。根据此种情况,提出一种核主分量分析与经验模式分解相结合的方法。该方法首先建立信号相空间,利用核主分量分析方法提取相空间的核主分量,然后利用投影逆过程将得到的核主分量逆向投影回原相空间,从而重建信号相空间。最后对重建的相空间所对应的信号作经验模式分解。此方法可以有效消除噪声和冗余对经验模式分解的影响,提高经验模式分解的适应能力保证分解的有效性,确保其能够分解出正确的基本模式分量。通过工程实例进一步验证了该方法的可行性。

Empirical mode decomposition based on kernel principal component analysis with application

If the signal-noise ratio of a signal is smaller, its intrinsic mode functions (IMFs) can not be decomposed correctly by using empirical mode decomposition(EMD), one of them may be a spurious IMF. Aiming at this case, kernel principal component analysis (KPCA) and EMD method were combined. Firstly, a signal phase-space was established, the kernel principal component of the signal phase-space was extracted using KPCA, and then, the phase-space was reconstructed by using the adopted kernel principal component through the inverse process. Finally, the recon-structed signal was decomposed using EMD method. The noise impact on EMD could be effectively eliminated by this way, so the decomposition ability of EMD was improved to ensure the validity of EMD in all cases and the correct IMFs were obtained. The effectiveness and feasibility of this method was verified further with actual examples.