Research On The Key Technology Of High-resolution Array Signals DOA Estimation

Direction of arrival(DOA)estimation is an important research problem in the array signal processing,which is widely used in a variety of practical applications,including missile-borne passive array direction finding,speech signal processing,5G communication,and radio astronomy.With the rapid development of modern information technology,the DOA estimation is facing increasing complex signal environment.Higher estimation accuracy and angular resolution and lower computational complexity are often required in practical applications,and there exist some specific scenarios that we need the proposed algorithms have the ability to achieve two-dimensional DOA estimation and undersetermined DOA estimation.In addition,the requirement of DOA estimation in demanding scenarios of low signal-to-noise ratio(SNR)and much limited snapshots have emerged in various areas.Therefore,this thesis mainly study the DOA estimation problem of coprime array in the radar signals processing,quasi-stationary signals in the speech signal processing and broadband non-stationary LFM signals based on the time-frequency analysis.For the DOA estimation problem of coprime array,this thesis firstly study the solving-ambiguity-based method according to the conventional coprime linear array,and proposes a novel DOA estimation algorithm based on the Multiple InvarianceEstimation of Signal Parameters via Rotational Invariance Technique(MI-ESPRIT)and a lookup table method.The MI-ESPRIT algorithm makes full use of the subarray elements,which improves the angle measurement precision.In addition,the proposed algorithm does not require the spatial spectrum search and uses a lookup table to solve ambiguity,which reduces the computational complexity.By using the relationship between the signal subspaces of two subarrays,the algorithm can avoid the matching error when multiple signals exist.Simulation results verify the effectiveness of the proposed algorithm.Then,this thesis studys the virtualization array sensor method based on the augmented coprime linear array,and proposes a more excellent DOA estimation algprithm based on the matrix completion theory.By constructing the selection matrix,we can directly operate on the covariance matrix of the continuous virtual array elements,without performing the covariance matrix vectorization and the spatial smoothing operation.In addition,we can successfully avoid the noise term and the error introduced due to a finite number of snapshots by constructing the error-constrained minimization function.After performing the low rank matrix completion operation,we can obtain the covariance matrix corresponding to the complete virtual array elements,then the DOA estimation can be realized based on the ESPRIT algorithm.The proposed algorithm further enhances the degrees of freedom(DOFs)offered by the co-prime,and the virtual array aperture is increased because the“holes” in the virtual array are filled based on the matrix completion,so the DOA estimation accuracy of the algorithm is also greatly improved.In order to improve the DOA estimation performance of quasi-stationary signals(QSS)at low signal-to-noise ratio and/or small snapshots,we propose novel DOA estimation algorithms of QSS in the context of sparse representation framework.Firstly,we study the DOA estimation of QSS based on the augmented coprime linear array.By constructing the received signal model of the augmented coprime linear array,we can make full use the advantages provided by the augmented coprime array in the virtual array domain,i.e.,high DOFs and large array aperture.In addition,we estimate the DOAs of QSS based on the sparse reconstruction method.Compared with existing mthods,the proposed algorithm can not only greatly increase the DOFs of DOA estimation,but also improve the estimation accuracy.Then,we study the DOA estimation of QSS based on the uniform circle array(UCA).By constructing the two-dimensional(2-D)over-complete basis of azimuth angle and elevation angle in the context of on-grid sparse representation framework,the 2-D DOA estimation problem of QSS can be conberted into an error-constrained convex optimization problem,which can be directly solved by CVX solver.However,the computational load of the CVX solver is unaffordable when the number of sensors is large.In order to reduce its computational complexity,the alternating direction method of multipliers(ADMM)solving mthod for DOA is exploited to solve the above error-constrained optimization problem.Based on dual decomposition and augmented Lagrangian,the convex optimization problem is divided into several de-coupling sub-problems so that the optimal solutions could be rapidly obtained by parallel computing.Therefore,the ADMM-based method greatly improves the computational efficiency.Since we make full use of the joint sparse representation of UCA under the sparse reconstruction framework,so the proposed algorithm has better DOA estimation performance,and the azimuth angle and elevation angle of QSS can be estimated simultaneously.Finally,considering the model mismatch problem in the on-grid sparse representation,we study the DOA estimation of QSS based on off-grid sparse Bayesian learning method.By describing the variables of the reconstructed model from a Bayesian perspective,an expectation-maximization iteration method is developed to estimate DOAs of QSS.Since we construct the off-grid model based on a linear approximation method with the first-order Taylor expansion,the modeling erroring caused by the basis mismatch is alleviated to some extent,so the estimation performance of QSS is greatly improved.DOA estimation performance of broadband non-stationary LFM signals may degrade substantially when LFM signals are spectrally-overlapped in time-frequency(TF)domain.In order to solve this problem,the single-source TF points selection algorithm based on STFT(WVD)and Hough transform is studied in this thesis,.Firstly,according to the empirical threshold value,the noise-term TF points in TF domain are filtered,respectively,and the cross-term TF points are also removed in the WVD.Then,the signal intersections in TF domain can be solved based on the Hough transform.By removing multiple-source TF points at intersections according to the empirical threshold values,we can get single-source TF points set.Based on the Euclidean distance operator,the single-source TF points set belonging to each signal can be obtained according to the property that TF points of the same signal have same eigenvector.Finally,the averaged STFD matrix of each signal is constructed and DOA estimation is realized based on the MUSIC algorithm.In this way,the proposed algorithm can resolve the TF non-disjoint LFM signals because it can automatically select single-source TF points set of each signal.In addition,the proposed algorithms have better DOA estimation precision and angular resolution,and can achieve underdetermined DOA estimation.When multiple LFM signals intersect at one point,the proposed algorithm in the STFT can still achieve DOA estimation.Numerical simulations and microwave anechoic chamber experiments demonstrate the validity of the proposed method.