Research On Parameter Estimation Of Frequency-hopping Signals In Impulse Noise

In recent years, due to its excellent anti-interference performance, low probability of interception and detection, outstanding compatibility, and many other advantages, Frequency-Hopping(FH) signals not only have been received wide attention of scholars both at home and abroad, and have been widely applied in military and civilian communications. Effective parameter estimation is of great importance to guarantee accurate information transmission and becomes a hotspot in the study of FH communication. Lots of studies show that many types of clutter interference or actual noise in radar, earthquake, biological engineering, obey a stable distributed and have significant peak pulse waveform and thicker tail of the probability density function and they do not have limited second order moments and high order moments. Under this kind of noise, the performance of conventional signal processing method based on Gaussian model degrades or even failures. Therefore, in the a stable distributed noise background, study on practical FH signals parameter estimation methods is of great significance for the research of FH communication. Parameter estimation of FH signals in a stable distributed noise is analyzed and studied, the main achievements are as follows:1. The choice of STFT window width in the traditional time-frequency analysis method is studied. To address this issue, an improved method based on Renyi entropy is introduced to evaluate the performance of time-frequency analysis, which is utilized to choose the minimum entropy values to adjust the window width, so as to gain the best window function. Compared with the traditional STFT, the Renyi entropy based STFT method combined with the fractional lower order statistics, can estimate hopping cycle accurately under the a stable distributed noise, what’s more, the estimation performance of hop timing and hopping frequency can be improved.2. The suppression of cross terms in quadratic time-frequency analysis methods is studied. In view of the problem that conventional non-linear time-frequency analysis methods in FH signals processing suffer from the effect of serious cross terms and parameter estimation precision reduce problems, the position of cross terms and auto terms are analyzed briefly, and then a Radially Gaussian Kernel(RGK) time-frequency analysis method is introduced in this paper. It selects adaptively optimal low-pass Gaussian kernel function to suppress the cross terms which are located away from the origin, and to concentrate the auto terms which are centered at the origin of the ambiguity plane. The experimental results show that the method has good time-frequency resolution and nice parameter estimation performance.3. To restrain the impulse noise, a method based on generalized Cauchy distribution is studied. A weighted maximum-likelihood generalized Cauchy(WMGC) filter is put forward. Combined with the principle of reliability, weighted maximum likelihood estimation theory is used to get the best sample value. The best weights are calculated according to the cost function minimum criterion, so as to select samples which are the closest to the expected value. Then, combined with the RGK time-frequency analysis method, the WR(WMGC-RGK) method is proposed to estimate the FH signals parameters in the ? stable distributed noise. Simulation results show that compared with the time-frequency analysis methods based on the fractional lower order statistics as well as the Myriad filter, WR method has better performance for FH signals parameter estimation and it is robust to the ? stable distributed noise.