The Research Of Signal Processing Method Based On The Empirical Mode Decomposition

Empirical Mode Decomposition/Hilbert Spectrum (EMD/HS) was first proposed by N E. Huang in 1998, which can decompose the nonlinear and non-stationary signals, then gives a better understanding of the physics behind the signals. As is well known, both wavelets and Fourier spectral analysis expand signals into a set of basis functions which are defined by the underlying method. Fourier transform can obtain a better resolution in the frequency domain, but not in the time domain. Wavelets can obtain a better resolution simultaneously in the time and frequency domains; but diminishing frequency-dominant range will result in the decrease of the wavelets of high-level detail components. It is disadvantageous to analysis results. EMD is based on the local characteristic time scale of the data. Compared with wavelets, EMD has all advantages of wavelets, dispels their fuzziness and unclearness, so can accurately reflects physics characteristics of the primitive signals.In this dissertation, the main research contents are as follows:1. The dissertation first reviews the EMD time-frequency analysis method, compares the EMD with some traditional time-frequency analysis methods, and gives the physics characteristics of Hilbert/Huang time-requency spectrum (HHT). Subsequently, the properties of EMD are summarized, and the pivotal problems are presented in the EMD algorithm. Finally, in terms of the 1-D EMD algorithm, the 2-D EMD method is given.2. In order to solve the primary problems in the former section, the deficiency of the konwn method is analyzed, then a new method of EMD edge problem based on wavelet-Kalman filtering hybrid forecast is proposed. The new method makes use of the real-time and recursive properties of Kalman filtering and multiscale analysis properties of the wavelets, and combines them to solve edge problem by data extension. 3. Due to the influence of noise in edge problem, the second generation wavelets are introduced in the dessertation, which have powerful de-noise characteristic. After comparing the noise EMD with the de-noise EMD, and analyzing the influence of noise on the EMD, the edge problem can be solved by the second generation wavelets de-noise technique.4. The performance of the new method has been tested by both mathematical theory and simulations. Furthermore, it is clear that the new method can greatly improve the precision of EMD and the performance of time-frequency analysis.