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

Time-frequency analysis is highly efficient in processing the non-linear and non-statinary signals, more and more research has been put on this topic. Time-frequency analysis method based on Empirical Mode Decomposition (EMD) is a new analysis method of signals. After several years' development, the EMD method shows its advantage in non-statinary signal processing, and has important theoretical research value and broad prospects. The EMD method is deeply studied both in theory and application in this paper.Firstly, the definitions of instantaneous frequency, characteristic time scale and intrinsic mode function are illustrated; reseach has been done on the algorithm of EMD. Then, the meanings in physics of Hilbert/Huang time-frequency spectrum and marginal spectrum are analyzed.Secondly, the analysis of the EMD algorithm in the hotspot and solutions is presented. The fitting problem of the envelope line is discussed emphatically. Compared with the Hermit interpolation and the spline interpolation, it proposed a new algorithm to improve the spline interpolation. The simulation result proved that this method can eliminate the overshoot and undershoot phenomenon.Thirdly, the EMD method is applied to extract the trend in non-stationary signal. Compared with traditional method, EMD is an adaptive, easily-implemented method. It presented three kinds of classical trends as stimulations to verify the validity of EMD.Finally, the filter character which based on the EMD's time scale is studied; then EMD threshold method has been proposed to filter the noise of signal. Examples from the results of statinary and non-statinary signals de-noising are given to demonstrate the efficiency of this new method. As an application, the EMD threshold method is considerd to have the advance in Raman spectrum de-noising.