Parameter Estimation Of LFM Signals In Alpha Stable Distributed

Linear frequency modulation signals are typically non-stationary,which have the characteristics of long acting distance,low probability of interception and high resolution.They are widely used in underwater acoustic,radar,biomedical and geological exploration.Frequency slope and initial frequency are phase parameters of linear frequency modulation signals,which contain important information,so the research of parameter estimation methods of linear frequency modulation signals has broad prospects.The parameter estimation methods of linear frequency modulation signals are generally aimed at Gaussian noise environment.However,it is found that the background noise usually has non-Gaussian characteristics in practical engineering applications.This kind of noise does not obey Gaussian distribution,and the probability density function has a thick tail and has obvious pulse characteristics.The alpha stable distributed model can accurately describe the noise.In this complex noise environment,conventional signal processing methods based on Gaussian hypothesis are no longer applicable.Therefore,it is of great significance to study the parameter estimation method of LFM signals under alpha stable distributed noise.The performance of traditional parameter estimation methods degrades or even fails under alpha stable distributed noise.In this thesis,we construct two kinds of nonlinear amplitude transformation functions with attenuation and bounded monotone increasing properties respectively,and then propose a new method for parameter estimation of LFM signals under alpha stable distributed noise based on these two functions.In this thesis,the function characteristics of the two kinds of functions in the near zero domain are analyzed,and the theoretical derivation proves that the linear frequency modulation signal in the impulse noise has the bounded second order statistics after the proposed transformation,and only the amplitude changes,the slope of the frequency modulation and the initial frequency value remain unchanged.Perform LVD analysis on the noisy signal after NAT transformation,and the parameter estimation of linear frequency modulation signal can be obtained according to the peak coordinates in transform domain.The simulation results show that the proposed method does not need the prior information of noise and can accurately extract the signal parameter information under the condition of strong impulse noise,with good robustness.In recent years,correntropy theory has been widely used in the field of impulse noise suppression.Correntropy has inherent robustness,and its higher-order expansion of contains such information as the autocorrelation function of the signal,and its kernel function is generally Gaussian,which can suppress impulse noise to a certain extent.However,the traditional correlation entropy has the defects that it cannot be applied to complex-valued signal processing and strong impulse noise environment,and its high-order expansion only contains the even-order moment of the signal.In this thesis,constructs a Laplacian correntropy(LC)capable of handling complex signals,and further proposes a robust Laplacian correntropy(RLC)function.The RLC function can suppress the impulse noise with close amplitude,and its higher-order expansion contains the information of all moments of the signal.In addition,STR-FT time-frequency analysis method for pulse noise is proposed.The time-frequency distribution of a LFM signal is linear after STR-FT transformation,and the parameter information of the signal can be extracted by Hough transform.Simulation results show that the STR-FT can effectively suppress impulse noise,accurately estimate the parameters of noisy signals without noise prior information,and still has robustness in strong impulse noise environment.The L-estimation method is often used to suppress impulse noise,which is achieved by selecting appropriate weights.However,there are still defects in this method,such as lack of unified weight selection criteria and weak ability to suppress impulse noise.Moreover,the signals in practice to be measured are greatly affected by impulse noise.In addition,based on the square Cauchy mixture distribution(SCM)distribution,this thesis proposes an optimal L-SCM weighted model,which can characterize the relationship between the data of the signal to be measured and the influence degree of the pulse noise.Furthermore,combined with Fr FT,an L-SCM-Fr FT signal parameter estimation method is proposed.The simulation results show that the proposed L-SCM weighted estimation method has a good noise suppression performance,compared with the FLO-FRFT method based on fractional low-order theory and the My-FRFT method based on Myriad filter,the L-SCM Fr FT method proposed in this article has a stronger noise suppression ability.It has good robustness without obtaining prior information of noise.