Research On Robust Adaptive Beamforming Algorithm Based On Matrix Reconstruction And Steering Vector Estimation

Array Signal Processing is a branch of Signal Processing,which is widely used in various fields such as radar system,wireless communication and speech signal pro-cessing and so on.As the name suggests,it refers to taking several sensors as array elements and arranging them into a specific array according to the design requirements.Finally,the signal received by the array is processed and analyzed with algorithms.Ar-ray signal processing includes two main directions:Direction of Arrival(DOA)and Beamforming.Beamforming is essentially a kind of spatial filtering.Based on some certain math-ematical criteria,the beamforming weight vector is designed to meet the requirements and the beam pattern is formed related to the spatial information such as the angle,so that the main lobe is in alignment with the direction of the desired signal,and the interference signal just falls into the beam nulling,so as to realize the so-called spa-tial filtering.The famous standard Capon adaptive beamformers are proved to be the optimal beamformers in theory.However,the interference covariance matrix and the desired signal steering vector are unable to be accurately estimated due to various errors in the actual scene,so the performance is often unsatisfactory.In the past few decades,the robustness of beamforming has been the focus of research.Among many exist-ing algorithms,the beamforming algorithm based on matrix reconstruction can better reduce the estimation error of covariance matrix,and has strong robustness.In this article,the analysis and research are focused on the interference covariance matrix re-construction based robust adaptive beamforming algorithm.Two new beamforming algorithms are proposed,namely robust adaptive beamforming algorithm based on in-terference covariance matrix reconstruction and steering vector searching,and robust adaptive beamforming algorithm for coprime array based on interference power esti-mation via subspace,respectively.The specific content and main innovation points of this research work are as follows:1.The reconstruction of the covariance matrix is closely related to the estimation of the signal steering vector,so the accurate estimation of the steering vector is studied in this paper.First,the traditional Capon power spectrum is analyzed from a more general perspective,that is,to analyze the function of the Capon power spectrum by treating the vector as the variable,and verify the existence of residual noise through theoreti-cal derivation.Also,it is proved that the more accurate the estimation of the steering vector is,the greater the corresponding Capon power value is.The traditional Capon power spectrum is utilized to estimate the nominal steering vector,and a hyperplane is constructed in the angle interval of each nominal steering vector neighborhood.the vector perpendicular to the hyperplane is taken as the gradient vector.Then use the nominal steering vector as the starting point to search along the gradient vector in the vector space.The Capon power value is used as the measurement standard.The vector corresponding to the maximum Capon power value is chosen as the final estimate of the steering vector,and the corresponding Capon power value is recorded at the same time.Finally,based on the Capon power and the steering vector estimate,the interfer-ence covariance matrix is reconstructed.Compared with other methods,the proposed method can obtain a relatively higher output signal-to-interference-plus-noise ratio with the same level of computational complexity.2.The current research on robust adaptive beamforming methods based on coprime arrays is still in its infancy.Since the good properties of the coprime array have a certain degree of incompatibility with the beamforming itself,this article hopes to better com-bine the more mature algorithms based on the uniform linear array with the advantages of the coprime array.Therefore,a robust adaptive beamforming algorithm for coprime array based on interference power estimation via subspace is proposed.This method first expands the aperture to obtain the virtual sampling covariance matrix of the virtual uniform linear array.Then it uses the approximate projection based on the uniform lin-ear array to remove the expected signal component in the covariance matrix.Based on the semi-definite property of matrix,the power and direction of each interference can be estimated via subspace in the virtual domain.Finally,the interference-plus-noise covariance matrix is reconstructed according to the structure of the real coprime ar-ray.Compared with the existing mainstream coprime array beamforming methods,this method has a higher output signal-to-interference-plus-noise ratio and better robustness.