An empirical signal separation algorithm for multicomponent signals based on linear time-frequency analysis

The empirical mode decomposition (EMD) is a powerful tool for non-stationary signal analysis. It has been used successfully for non-stationary signals separation and time-frequency representation. Linear time-frequency analysis (TFA) is another powerful tool for non-stationary signal. Linear TFAs, e.g. short-time Fourier transform (STFT) and wavelet transform (WT), depend linearly upon the signal analysis. In the current paper, we utilize the advantages of EMD and linear TFA to propose a new signal reconstruction method, called the empirical signal separation algorithm. First we represent the signal with SIFT or WT. After that, by using an EMD-like procedure, we extract the components in the time-frequency (TF) plane one by one, adaptively and automatically. With the iterations carried out in the sifting process, the proposed method can separate non-stationary multicomponent signals with fast varying frequency components which EMD may not be able to separate. The experiments results demonstrate the efficiency of the proposed method compared to standard EMD, ensemble EMD and synchrosqueezing transform. (C) 2018 Elsevier Ltd. All rights reserved.