Research On Parameter Estimation Of Frequency-hopping Signals In Alpha Stable Noise

The frequency hopping (FH) technique is one of the spread spectrumcommunication modes, and has been widely used in the field of military and civilian onaccount of its strong anti-jamming, low probability of intercept and excellentnetworking capabilities. How to effectively realize FH parameter estimation attractsmore and more attention in modern communications.In traditional FH signal processing, the Gaussian model is used as backgroundnoise model for its excellent features, but it is not suitable in the actual noise or clutterenvironment, such as communication channel interference, atmospheric noise,electromagnetic disturbance etc.. These noises have a remarkable impulsivecharacteristic. Based on extensive research, the alpha stable distribution has beenproposed as an effective model to describe the non Gaussian pulse noise. In such abackground, the performance of the conventional processing methods based onGaussian distribution can be reduced or even disabled. Therefore, this dissertation isfocused on the FH signal parameter estimation in alpha stable noise environment.As a typical non-stationary signal, the traditional Fourier transform method is nolonger applicable for the FH signal. The time-frequency analysis method shows a goodpicture in the joint time-frequency domain and offers the change of frequency alongwith time, so it becomes a powerful tool for FH signal parameter estimation. Amongkinds of commonly used methods, the short-time Fourier transform (STFT) and theWigner-Ville distribution (WVD) are the basic time-frequency analysis methods. TheSTFT is widely adopted in practice because of its simple principle, low computationalcomplexity, good real-time capability and engineering feasibility; The WVD and thepseudo Wigner Ville distribution (PWVD) have higher time-frequency resolution thanthe STFT, but they are disturbed by the cross-term interference seriously. The smoothpseudo Wigner-Ville distribution (SPWVD) shows great performance in terms oftime-frequency resolution and cross term inhibition, but it is not satisfied in practicebecause of the large computation complexity and poor real-time capability. Byanalyzing and contrasting several time-frequency analysis methods, the STFT is chosenas the time-frequency analysis method.Conventional time-frequency analysis is a powerful tool for the FH signalprocessing, however, it fails to realize parameter estimation in alpha stable noise environment. Three improved Time-Frequency analysis methods are proposed to solvethis issue. One is the fractional lower order STFT and uniform DFT filter (FDSTFT),the other two are the STFT based on the Myriad filter (MYRSTFT) and the Meridianfilter (MSTFT) respectively, both of which are based on Maximum Likelihood (ML)estimation. Simulation results show that the proposed methods have better parameterestimation performance for FH signals than the STFT and the fractional lower orderstatistics based STFT (FLOSTFT) in alpha stable noise environment.