Radar Signal Parameter Estimation with Sparse Bayesian Representation Based on Zoom-Dictionary

Abstract

In this paper, a sparse Bayesian representation model based on zoom-dictionary is proposed to estimate the parameters of the complex radar signal. According to the traditional parameter estimation methods, the parameters in a signal can be estimated via a dictionary whose atoms defined on the discrete grid in the parameter space. Then, we use a Beta-Bernoulli prior to realize the sparse selection of the parameterized atoms in a given dictionary (i.e. Achieve the estimation of the parameters), where the dictionary grids can be iteratively zoomed in to estimate the parameters more precisely. The varaitional Bayesian (VB) inference algorithm is developed for the proposed sparse Bayesian model. Furthermore, based on the preliminary parameter estimation, we develop a weighted clustering method to further enhance the accuracy of the parameter estimation. Simulation examples demonstrate that the proposed algorithm can achieve much more accurate parameter estimation results.