Abstract:

To address the problem that the traditional timefrequency analysis methods based on the Wigner Hough transform (WHT) degrade severely in stable noise environment, an optimal L-Cauchy weighted (LCW) method based on the L-estimation theory which can effectively suppress this kind of noise is proposed. The 3En criterion is a kind of commonly used method to eliminate outliers, which can effectively restrain the outliers from the point of mathematical statistics. Combined with the Cauchy distribution, a method to restrain outliers based on the dispersion coefficient is proposed, and the optimal parameter value of α in the LCW method could be selected by numerical simulation. The parameters of noisy linear frequency modulation (LFM) signals can be estimated by the WHT method based on the LCW method. Simulation results show that the optimal parameter value of α consists with the proposed method for restraining outliers. Compared with the L-estimation, the fractional lower order statistics as well as the weighted Myriad filter based time frequency analysis methods, the proposed method has better performance for the LFM signal parameter estimation and it is robust to the α stable distribution noise.