Research On Near-field Source Location Based On Tensor Decomposition

The core of array signal processing includes beamforming technology,zero shaping technology and spatial spectrum estimation,which is often called direction of arrival(DOA)estimation and plays a key role in the field of source location.According to the distance between source and antenna array,it can be divided into near-field source signal and far-field source signal.The far-field source model contains only angle information,while the near-field source contains both angle and distance parameters.In this paper,starting from the establishment of the parameter estimation model,aiming at the more complex near-field sources,we use the tensor decomposition(TD)technology to estimate the parameters of the near-field sources in the white noise environment,and then achieve the target of near-field source location.The main work is as follows:1)The basic knowledge of matrix and tensor used in this paper is introduced.In this paper,the operation rules and the corresponding operation symbols of matrix are unified,the concepts of tensor and its rank are introduced,and the conditions for the uniqueness of tensor decomposition are analyzed.At the same time,the basic receiving model of near-field source signal is deduced.2)Through theoretical derivation,a parameter estimation method for DOA and distance of near-field single source in two-dimensional space based on uniform symmetric linear array is proposed.Based on the symmetry of the receiving array,the tensor models for angle and distance parameter estimation are derived and constructed respectively,and the CE algorithm is proposed to perform canonical multivariate decomposition(CPD)to complete the estimation.The method is simulated and compared with the traditional 2d-music method.3)Through theoretical derivation,a parameter estimation method based on uniform rectangular array for angle and distance of near-field single source in three-dimensional space is proposed.Based on the characteristics of manifold matrix and the symmetry of array,the tensor model for angle parameter estimation is derived and constructed,and the tensor decomposition is completed by using alternating least squares(ALS)to estimate the angle parameter.Then the distance parameters are estimated by the original received signal model.The effectiveness of the method is verified by computer simulation,and its performance advantage in low snapshot environment is discussed by comparing with other LS algorithms in the same scene.4)A parameter estimation method for angle and distance of near-field multiple sources in three-dimensional space is proposed.Using L-shape array model,a multi parameter tensor model is constructed by second-order cumulant.According to the conjugate symmetry of the model,the improved ALS algorithm is used to decompose the tensor twice,and the parameters are matched by the distance term.The estimation performance of this method is obtained by computer simulation,and its advantages and disadvantages and applicable environment are analyzed.5)The performance and application value of tensor decomposition method in the near-field source localization problem are discussed generally.The nature of the proposed method is discussed,and its limitations are explored.Suggestions are also made for the development of related research in the future.