Research On The Time-Frequency Analysis Method About Nonstationary Signals

Time-Frequency analysis is one of the research hotspots in signal processing in recent years, and it is powerful tools for analyzing the non-stationary signals. Because it provides the joint distributed information between time domain and frequency domain, we can know about the change of frequency along with the time clearly. In this paper, Time-Frequency analysis is regarded as the key research subject and some methods are introduced, such as Short Time Fourier Transform(STFT), Wigner-Ville Distribution(WVD) ,Wavelet Transform(WT) and so on.Firstly, a new technology-Hilbert-Huang Transform is analyzed systematically. Based on the existed theories, an improved method for the Hilbert-Huang Transform's core algorithm-empirical mode decomposition (EMD) is presented. And a novel method based on the improved EMD is provided to reduce the crossterms in Wigner-Ville Distribution. At the same time, pseudo component distinguishing method is presented too.Secondly, the simulation for several kinds of non-stationary signal based on STFT, WVD, WT and the improved Hilbert-Huang transform is given. Simulation results verify the validity of the improved Hilbert-Huang transform in time-frequency concentration and eliminating the crossterms, it has better effect than morlet wavelet spectrum, WVD and STFT.Lastly, put the improved Hilbert-Huang Transform in the use of analysis time-frequency characteristics of classical nonlinear equation system, non-stationary signal denoising and processing the variable speed parameter. Simulation results show the efficiency of time-frequency analysis method and its superiority in demonstrating the local features of the processed signal.