Research On Direction Of Arrival Estimation Algorithm For Mixed Sources

Direction of arrival(DOA)estimation,also referred to as directon finding technique,is an important research branch in array signal processing,which has been widely used in the field of military and civilian,such as radar,sognar,navigation and communication systems.Spatial spectrum estimation technique can resolve multiple incident sources in a beam accurately,and its good performace in general is based on the assumption of uncorrelated or low correlated sources.However,in the practical environments,due to various interferences and the scenario of multipath propagation caused by reflective surface,such as hills,clouds and buildings,there exist a large number of coherent sources.As a result,the coexistence of uncorrelated and coherent sources is very commom in the real direction finding system.When dealing with the coherent sources,the traditional spatial spectrum estimation algorithms for the uncorrelated sources would suffer from performance deterioration owing to the rank deficiency of array covariance matrix.Therefore,it is of great practical significance to investigate the DOA estimation algorithms for a mixture of uncorrelated and coherent sources.This dissertation focuses on the DOA estimation algorithms for mixed sources,aiming at proposing some novel solutions to some problems in the existing algorithms.The major contributions of this dissertation are described as follows:Firstly,for the purpose of reducing the computional complexity and extending the array aperture,we propose a real-valued DOA estimation algorithm for a mixture of uncorrelated and coherent sources.The utilization of the unitary transformation lays a foundation for the estimation of both uncorrelated and coherent sources in the real domain.In the implementation process of this algorithm,the oblique projection operator is employed to distinguish the uncorrelated sources from the coherent sources,then the DOAs of the uncorrelated sources and the coherent sources are estimated separately via the construction of real-valued augmented matrices,the real-valued rotational-invariant submatrices and Hankel matrix reconstruction.Besides,two augmented matrices are constructed during the estimation process,which helps to extend the effective array aperture.In addition,the fading coefficient estimation problem is transformed into a modified constraint quadratic minimization problem by adding penalties,which guarantees the stability of the solution.Simulation results demonstrate the effectiveness of the proposed algorithm.Secondly,we study on how to improve the DOA estmation accuracy and angular resolution by taking full advantage of the polarization diversity of the impinging sources,andpropose two joint DOA and polarization estimation algorithms based on vector sensor arrays.The first one is a 1-D joint DOA and polarization estimation algorithm using a dual-polarization vector sensor array.In this algorithm the uncorrelated sources are separated from the coherent sources on the basis of the modulus property of eigenvalues,then the DOA and polarization parameters of both uncorrelated and coherent sources are estimated respectively by using the rotational-invariant submatrices,the reconstruction of array manifold matrix and Hankel matrix,and the least square theory.However,in accordance with the spatial Nyquist sampling theorem,the inter-sensor spacing of the array adpotted by this algorithm is limited to a half-wavelength.The second one is a 2-D DOA and polarization estimation algorithm using an L-shaped sparsely-distributed vector sensor array.A series of L-shaped sparsely-distributed vector sensor array are developed,with which we realize 2-D DOA and polarization estimation.In the process of implementation of this algorithm we still separate the uncorrelated sources from the coherent sources using modulus property of eigenvalues,but this 2-D algorithm achieves the refined DOA estimation by combing with the coarse estimates and accurate estimates with cyclical ambiguity.Moreover,the inter-sensor spacings are allowed beyond a half-wavelength,which extends the effective array aperture,and the estimation accuracy is improved accordingly.Simulation results vertify the favorable performance of the two proposed algorithms.Finally,we focus on how to realize the DOA estimation for a mixture of uncorrelated and coherent sources without resorting to decorrelating operations and source number estimation algorithms,and propose a DOA estimation algorithm for mixed sources under the framework of sparse Bayesian learning.To encourage sparse solutions,we introduce a new three-layer sparse-encouraging prior,referred to as Gauss-Exp-Chi2 prior.Based on this prior,the three-layer sparse Bayesian model is established,and then the relevant model parameters are estimated via the variational Bayesian inference.By calculating the source power spectra,the source number estimation and DOA estimation are required.Compared with the existing subspace-based DOA estimation algorithms for a mixture of uncorrelated and coherent sources,the DOA estimation performace of this algorithm does not depend on the accuracy of decorrelating and source number estimation results,which improves its reliability.Simulation results show the effectiveness of the proposed algorithm.