Research On DOA Estimating Approaches Based On Compressive Sensing

Compressive sensing(CS)theory is the latest technology in the past decade,it uses the characteristic which the projection of signal in a certain space is sparse or compressible to reconstruct the sparse signal accurately from a small amount of sampled data.Therefore,the algorithm of estimating direction of arrival(DOA)sources based on compressive sensing perception makes huge contribution to engineering domain.This thesis aims to use the CS theory to enhance the direction of arrivals estimation performance,the details are as follows:Firstly,introducing the basic theory of CS and several traditional DOA estimation methods.Carrying out several simulations to verify the effiectiveness of these algorithms,and analysing the performance of these algorithms,laying the foundation for following research for the estimation algorithm based on CS theory.Secondly,focused on the far-field and narrowband signal in the circumstance of uniform linear array,two kinds of super-resolution DOA estimation algorithms are introduced which is mixed norm constraint methods including 1l-SVD,weighted 1l-norm minimization algorithm,Sparse Representation of Covariance Vectors(SRACV)algorithm,and Bayesian compressive sensing-based DOA estimation algorithm.The effectiveness of these algorithms are verified by some simulations,and the performance these algorithms and the traditional estimation algorithm is compared.Thirdly,focused on the 2D-DOA estimation algorithm.We briefly introduced the 2D-DOA estimation methods extended from the classical 1D-DOA estimation methods and simulations had been done to evaluate the performance of these methods.Then,we introduced a novel approach for 2D estimation based on the separable observation model using signal snapshot.Separating the over-complete dictionary into two individual subdictionary is the key idea of this method so that the computational complexity is reduced heavily.But it needs to choose two regularization parameters and they are difficult to be obtained accurately,so the performance of this method would degrade when the SNR is low and the scale of planar array is small.And the simulations had been carried on to prove the result to this method.Finally,we proposed a novel approach for 2D-DOA estimation based on the weighted 1l-norm minimization method and multitask Bayesian compressive sensing framework using this new observation model.The key idea of this method is separating the estimation into two steps,the weighted 1l-norm minimization method is used to enhance the estimation accuracy of elevations firstly so that have few influence to estimate the sparse signal,and there would be a guideline to choose the regularization parameters,then multitask Bayesian compressive sensing method is used to estimate the sparse signal as the noise distribution is difficult to derivate.The simulations had validated good performance in the low SNR and small scale planar array circumstance.