经验模式分解回顾与展望

Review and prospects of empirical mode decomposition

经验模式分解EMD打破了Fourier变换、小波分解等传统数据分析方法需要预先设定基函数的局限,是一种完全由数据驱动的自适应非线性非平稳时变信号分解方法,可以将数据从高频到低频分解成具有物理意义的少数几个固有模态函数分量和一个余量。首先介绍了原始EMD方法的原理和算法;接着,总结归纳了EMD当前的研究现状,分析了EMD存在的端点效应、模态混叠、运行速度问题及其在二维情况下的问题并对国内外学者解决这些问题的方法进行了概述和比较;最后结合EMD研究存在的难题指出了EMD进一步研究与应用的发展方向。

Empirical Mode Decomposition （EMD） is a totally data-driven and self-adaptive decompo- sition algorithm that is used to analyze nonlinear, non-stationary and time-varying signal, it breaks out the limitation of needing of presetting basis function for traditional data analysis method like Fourier transformation and wavelet decomposition, and it can decompose a signal into a few intrinsic mode func- tion components with physical meaning and a residue from high-frequency to low-frequency. Firstly, the principle and algorithm of the original EMD method are introduced. Secondly, we present an overview of the current development of EMD and analyze EMD＇s existing end effects, mode mixing, running speed problems and the problems that appeared when the original data are two-dimensional and compared re- searchers＂ solutions to these problems. Finally, combined with its problems, several directions of further research and application are pointed out.