Non-stationary Data Processing Based On Hilbert-Huang Transform

Analyzing and processing non-stationary signal is a heated topic and difficult issue in information processing domain. In 1998, Norder E. Huang et al. presented a new method for analyzing and processing non-stationary data that was called Hilbert-Huang transform (HHT). HHT is an adaptive signal analysis method, whose key part is empirical mode decomposition (EMD). Since HHT was presented, it has widely been applied to many engineering domains, such as earthquake signal analysis, mechanical vibration engineering, fault diagnosis and biomedicine engineering, etc, and has exhibited unique advantages. However, the theory of HHT is still imperfect, there have been many issues deserving more research. Therefore, the in-depth study of HHT technology bears not only realistic academic interest and use value, but also long-term social effect and economic significance.On the basis of in-depth researches on HHT, this thesis mainly covers the following research production:Firstly, aimed at the end effect in EMD process, two improved EMD algorithms based on combining extrapolating extrema with mirror extension and based on envelope extension method using AR model are respectively proposed. The capabilities of the presented EMD algorithms and the typical EMD algorithms are compared in the simulation experiments. The experimental results demonstrate that both the presented methods can better restrain the end effect.Secondly, in order to improve the relatively limited noise-filtering ability of EMD, the effective HHT processing approach by combining low-pass filtering with EMD for low-frequency signal with noise is presented in this thesis. Corresponding simulation experiments with the proposed method are carried out. Furthermore, some measured non-stationary liquid-level echo vibration data being processed, results demonstrate that the presented method is very effective.Finally, under the guidance of data prediction principle, the non-stationary data prediction method based on HHT is proposed. Typical signals with different trends being predicted in the simulation experiments, the experimental results validate that the presented approach is feasible and effective.