Research On Wideband Radar Signal Parameter Estimation Algorithm Based On Distributed Optimization

There is a problem of insufficient estimation accuracy and slow processing speed in the traditional wideband radar signal parameter estimation method.In order to estimate the parameters of wideband radar signal accurately and quickly,and to solve the large-scale optimization problem of reconstructing wideband FM signal accurately,this paper proposes a variety of wideband radar signal parameter estimation methods based on convex optimization and distributed algorithm.1.In order to improve the accuracy of parameter estimation of wideband FM signal,two parameter estimation methods based on the convex optimization basis pursuit algorithm for different dictionaries are proposed.To solve the problem of parameter estimation of nonlinear FM signal,a weighted multi-window dictionary is proposed.It solves the shortcomings of fixed window width which can't improve the resolution of time-frequency analysis.The parameter estimation accuracy of the nonlinear frequency modulation signal is improved.For the linear frequency modulation signal,the four-parameter Chirplet dictionary is used to fit the linear frequency modulation signal.The simulation results show that the algorithm can achieve high-precision signal reconstruction.2.Aiming at the shortcoming of slow speed in parameter estimation of traditional wideband LFM signal,three estimation methods based on alternating direction multiplier method are proposed.The group sparse group Lasso model,BP basis pursuit model and atomic norm soft threshold model are used to model the parameter estimation problem.The problem structures of the three models are decomposed and optimized by the alternating direction multiplier method.Finally,the fast reconstruction and parameter estimation of LFM signal are realized by alternating direction updating.3.To solve the problem of parameter estimation of multi-component LFM signal,a method of parameter estimation based on manifold clustering is proposed.Firstly,the signal is reconstructed and de-noised by GL-ADMM algorithm.Then the time-frequency distribution is obtained by rearranging the smoothed-pseudo-WV distribution.For time-frequency distribution results,spectral curvature clustering and curvature constrained path clustering are used to separate signals and estimate parameters.The simulation results show that manifold clustering algorithm can achieve the separation and parameter estimation of multi-component FM signals.