Study On 2-D Angle Of Arrival Estimation In The Sparse Bayesian Learning Framework

Direction of arrival(DOA)estimation is one of important issues in array signal processing.It mainly studies the ability of antenna arrays to accurately estimate signal spatial parameters and has been widely used in various military and civilian fields such as radar,sonar and wireless communications.Generally,DOA estimation revolves around the one-dimensional DOA estimation problem.A large number of excellent superresolution subspace-based estimators have been proposed,such as MUSIC and ESPRIT.In practical applications,the propagation direction of electromagnetic signals in threedimensional space is characterized by the two-dimensional angle of arrival(2-D AOA),which means that it is more realistic to do 2-D AOA estimation.If a one-dimensional DOA subspace-based method is extended directly to a 2-D AOA estimation problem,it usually results in a series of problems,e.g.,increased computational burden,low efficiency,and parameter matching.Recent studies have shown that sparse representation theory can provide a new way to obtain the super-resolution performance for DOA estimation,which is different from traditional subspace-based methods.DOA estimation methods using sparse representation theory have potential to deal with a 2-D AOA problem with super-resolution performance and can overcome the above-mentioned adverse effects coming from subspace-based methods for 2-D AOA estimation.This study uses a typical sparse representation framework,i.e.,sparse bayesian learning(SBL)to handle 2-D AOA estimation,which includes:1.A 2-D AOA estimator using SBL that can achieve automatic matching of azimuth and elevation angles is studied based on a dual parallel array structure model.The effectiveness of the method is demonstrated as comparied with the 2-D MUSIC algorithm by simulations.In order to achieve better results,the sparse vector model reconstructed from array output covariance matrix with greater degrees of freedom and higher array output signal-to-noise ratio is combined with the SBL algorithm for 2-D AOA estimation with automatic angle matching.The simulation results are compared with the method based on the original data model in terms of estimation accuracy,running time,and resolution,which proves that the 2-D AOA estimation algorithm based on covariance matrix reconstruction has higher accuracy.2.For studying the effect of different covariance matrix reconstruction methods on estimation performance,a covariance matrix signal model with block sparseness characteristics is introduced and is combined with block SBL(BSBL)for 2-D AOA estimation with automatic angle matching.Simulation results are compared with the SBL 2-D AOA estimators based on the original data model and the sparse vector model,which shows that the algorithms based on different covariance matrix reconstructions and SBL have a large difference in running time,but have very close estimation accuracy.3.In order to reflect the performance gap between the SBL algorithms and the norm optimization methods,a fixed-point iterative algorithm that needs not call the optimization toolkit is introduced based on a covariance matrix reconstruction model,and is compared with the SBL 2-D AOA estimator by simulations.4.To handle the off-grid effect in sparse DOA estimation,we use the statistical maximum likelihood estimator to obtain a higher accuracy solution,and compares it with the off-grid method based on Taylor series tools.