Research On Parameter Estimation Of Frequency-hopping Signal In Complicated Noise

As an important spread spectrum communication technique,frequency hopping(FH)spread spectrum has been widely utilized in modern communication systems.In non-cooperative FH communication,it is difficult to capture the communication content for the receiver because of the lack of relevant FH parameters.Therefore,the blind parameter estimation of FH signals has aroused general interest in non-cooperative FH communication technologies.Gaussian distribution has been taken to model the background noise in most of the classical statistical signal processing methods.However,lots of researches show that a great amount of actual measured data in complex noise environment appears with notable impulses,which presents typical non-Gaussian behavior.It is proved that ?-stable distribution can be utilized for modeling these situations,which means that Gaussian hypothesis based signal processing methods are faced with severe performance degradation,and new parameter estimation schemes are needed to be presented.For the diversity of communication channels,this thesis studies the parameter estimation of FH signals in both Gaussian and ?-stable noise environment.The major work is outlined as follows:1.In conventional time-frequency analysis based parameter estimation of FH signal under Gaussian noise,the suppression of cross-terms in Time-Frequency Distribution(TFD)by kernel functions always leads to the decrease of the time-frequency concentration,which is adverse to signal parameter extraction.To deal with this problem,a kind of Sparse Time-Frequency Distribution(STFD)based on the property of FH signals' TFD has been proposed by combining Cohen's class of TFD and the Compressed Sensing(CS)theory.STFD can not only restrain cross-terms effectively,but also has a high time-frequency concentration.Simulation results show that the proposed STFD based parameter estimation method for FH signals demonstrates better performance compared with conventional ones in Gaussian noise.2.In ?-stable distribution environment,most of the statistical models adopted in classical robust filtering are special cases of ?-stable distribution(such as Cauchy distribution and Meridian distribution,etc.),which confines their performance of impulse removal.Based on the robust filtering theory and the approximate symmetric ?-stable(S?S)distribution model,a new weighted maximum-likelihood(ML)approximate S?S(WMAS)filter is proposed in this thesis.The statistical model employed in the WMAS filter is an approximation of the S?S distribution probability density function(PDF)that uses power function,which can well describe the heavy tail of the S?S distribution PDF.This characteristic enables the WMAS filter to suppress the ?-stable noise adaptively and gain good robustness.3.A new ML-based method for FH signals parameter estimation is proposed.First,the complicated FH signal model is simplified by its frequency and envelope,and the ML-based signal frequency estimation frame is established based on the Cauchy distribution;a new iterative method is then derived to achieve effective estimation of the frequency parameter in ?-stable noise.Meanwhile,exploiting the short-term stability of FH signals,a sliding time function is designed by the window duration increasing optimization(WDIO)algorithm,which benefits the extraction of FH frequency variation curve.Then the hopping time and the hopping period can be estimated through the curve.Simulations show that the proposed method is more robust than the time-frequency analysis based on fractional lower order statistics and Meridian filtering while ensuring the effective estimation of the FH signal parameters.