Fig.2 An example of a standard EMD for the 1s-MMG signal (500 sample points).

The standard EMD decomposes a signal x(𝑡) into N sets of frequency- and amplitude-modulated signal components by employing an iterative process called the sifting algorithm, such asx(𝑡)=Σ𝑐𝑖(𝑡)𝑁𝑖=1+𝑟(𝑡),𝑡=1,…,𝑛,𝑖=1,…,𝑁 (1)where the series {𝑐𝑖(𝑡)}𝑖=1𝑁 is called the intrinsic mode functions (IMFs), which represent the intrinsic oscillations present in the raw signal, and 𝑟(𝑡) is a residual trend of the lower frequency (Fig. 2). In brief, the IMF is extracted from a signal by subtracting the average of the upper and lower envelopes, where the envelopes are respectively generated by concatenating all local maxima and minima.

标准经验模态分解（Empirical Mode Decomposition，EMD）通过采用一种称为筛选算法（sifting algorithm）的迭代过程，将信号x(t)分解为N组频率和振幅调制的信号分量，例如x(𝑡)=Σ𝑐𝑖(𝑡)（i=1到N）+𝑟(𝑡)，𝑡=1,…,𝑛，𝑖=1,…,𝑁（1）。其中序列{𝑐𝑖(𝑡)}（i=1到N）被称为本征模态函数（Intrinsic Mode Functions，IMFs），代表原始信号中存在的本征振荡；𝑟(𝑡)是低频的残余趋势（图2）。简言之，IMF通过减去上下包络的平均值从信号中提取，其中包络分别由所有局部极大值和极小值连接生成。