Time-varying frequency-shifting signal-assisted empirical mode decomposition method for AM–FM signals

A time-varying frequency-shifting signal-assisted empirical mode decomposition (EMD) method is presented in this paper. For AM–FM signals, if the extrema density (average number of extrema per unit length) is higher, the frequency is higher approximately, and vice versa. The assisted signal in this method is directly derived from the local extrema using the least square method, and avoids the difficulty that masking signals are not adaptively examined so that the masking signal may not always enable the EMD to generate true mono-component intrinsic mode functions (IMFs). Therefore, the constructed assisted signal is adaptive and suitable for the analysis of frequency-varying components. Compared with the traditional EMD, the method mainly solves the following problems of the traditional EMD: the traditional EMD fails to separate the modes whose frequencies are time-varying and lie within an octave, and fails to remove the time-varying frequency mode-mixing. The results of three experiments are given to show our method's performance and efficiency.