Theroretical Study And Application Of Time-Frequency Analysis Method Based On Empirical Mode Decomposition

Time-frequency analysis is one of the top interests in signal Processing and feature extracting of non-stationary signal, and more and more research has been put on this topic. And time-frequency analysis based on Empirical Mode Decomposition (EMD) is a new two-step time-frequency analytic method to process nonlinear and non-stationary signal. The key step of this method is EMD method with which any complicated data can be decomposed into a finite number of Intrinsic Mode Functions (IMF). Using Hilbert transform to those IMF components can obtain meaningful instantaneous frequency, the final presentation of this results is an energy-frequency-time distribution, designated as the Hilbert spectrum.Firstly, this paper discussed the definitions of instantaneous frequency and IMF, then the EMD algorithm was presented. After that, meanings in physics of Hilbert/Huang Transform spectrum and marginal spectrum have been analyzed, and some problems which may effecting the EMD algorithm have been analyzed.Secondly, based on these, the reasons which influenced on the accuracy of EMD have been analyzed, and the reasons which generated by lower sampling frequency for signals and end effect have been improved. In order to increase the sampling frequency, an interpolation manner which was used to add the sampling points was presented; then, in order to weaken the end effect in EMD algorithm, data extension technologies have been proposed. Algorithms of mirror based extension and neuron network based data extension technologies have been discussed, meanwhile their limitations have been pointed out, too. For this, an end extremum envelope extension algorithm has been presented. Simulation results show that end extremum envelope extension is an efficient and excellent method to eliminate end effect.Thirdly, the paper studied the self-adaptive capability of EMD method,time-frequency distinguish ability of the EMD time-frequency spectrum and the filter character which based on the EMD's time scale; then, based on the EMD's time scale filter character, filter manner of EMD's time scale based on related degree have been proposed. Simulation results show that this filter manner is effective and feasible.Finally, time-frequency analytic method based on EMD is used to extract of signal's transient feature and to extract of signal's trend component to demonstrate the efficiency and superiority of this new time-frequency analytic method.