Modified S-Transform And Its Application To Instantaneous Frequency Estimation

Instantaneous frequency estimation is an important method to analyze non-stationary signals, and it is widely used in military and civil fields. It has important practical significance in the field of non-stationary signal analysis to study the time-frequency analysis tools with high energy conversion and high frequency resolution. In this thesis, a modified S-transform, which is based on the standard S-transform with linear and multi-resolution characteristics, is studied. And apply this modified S-transform to instantaneous frequency estimation. The main research contents of this thesis are as follows:Firstly, two kinds of conventional time-frequency analysis methods which are the linear time-frequency representation and the quadric time-frequency distribution are studied. The performance analysis shows that, the S-transform has a completely reversible transformation, which has the advantages of preserving the phase information as well as STFT and the variable resolution of WT, while it can effectively remove the cross term interference due to its linear. Compared to other common time-frequency analysis methods, the S-transform has a good applicability and potential in the process of dealing with non-stationary signals.Secondly, for the problems of the S-transform, such as low frequency resolution is obtained, the aggregation is not high enough, and a distortion exists in the high frequency area because of the noise, several kinds of generalized S-transform based on changes of the window function are studied. The multi-window generalized S-transform is emphasize studied. Compared to the standard S-transform, the modified S-transform has higher time-frequency energy concentration and higher accuracy of IF estimation. In addition, the multi-window Stransform has much higher concentration and resolution in time-frequency domain than other conventional generalized S-transform, and it can suppress the noise.Finally, in order to apply more complex signals and higher requirements, a new adaptive Stransform method based on a corresponding concentration measure(CM) is proposed. By adding the optimal control parameters to the original S-transform, the control of the window width is realized and the optimal parameters of each frequency are obtained according to the proposed optimization concentration measure. This method can effectively analyze the multi component signal in different frequency band and adaptively adjust the time-frequency resolution parameters according to the time varying characteristic of the signal. Simulation experiments show that the proposed method has better time-frequency concentration and more accurate IF estimation for multi-component non-stationary signals.