Study On DOA Estimation Based On Distributed Subarrays And Spatial Sparsity

The modern radar is faced with severe environments,such as low SNR?Signal-to-Noise Ratio?,limited snapshots,which is due to the wide applications of the high-speed targets,and the increase of the electromagnetic emitters.For the detection and tracking of the distant targets,the SNR is low.However,the requirement of DOA?Direction-of-Arrival?estimation accuracy is improved in modern war for accurately attack the targets.The sparse array and the spatial sparse property of the received signal of the sensor array can be used to solve the above problems efficiently.The distributed array geometry can be applied to radar,which can possess large aperture with small antennas,and thus improve the DOA estimation accuracy and resolution.The sparse recovery methods for DOA estimation have good performance in low SNR and small snapshots number.However,the sparse recovery methods have a large computation cost.This dissertation focuses on the distributed array radar and the sparse recovery methods for DOA estimation.The main contents are listed below:1.Optimize the sub-array of the distributed array radar.The nested array is used as the sub-array of the distributed array radar,and the degree of freedom of the sub-array is improved.The geometry of the distributed nested array is given.According to the geometry of the distributed array radar,the multi-resolution ESPRIT?Estimation of Signal Parameters via Rotational Invariance Technique?algorithm is used to estimate DOA when the number of the targets is small.The MAP?Maximum a Posteriori?method is used to analyze the performance of DOA estimation.The SNR threshold and the baseline threshold of the distributed array radar are analyzed.The computer simulation results is in accordance with the theoretical analysis.Simulation results show that the proposed method can decrease the SNR threshold because of the use of the middle estimation.When the targets number is larger than the number of the antennas of the sub-array,the covariance matrix is first vectorized to improve the degree of the freedom,the spatial smoothing unitary dual ESPRIT algorithm is then used to DOA estimation.Simulation results show the proposed method has high DOA estimation accuracy and resolution.2.Study on the multi-frequency distributed array.The holes of the co-array of the distributed array can be filled with the covariance matrix at the other frequency,then we can obtain a consecutive co-array without holes.The requirement of the additional frequency and the sensors which is needed to work at multi-frequency is analyzed.The spatial smoothing MUSIC?Multiple Signal Classification?is used to obtain the final DOA estimation with the consecutive co-array.Simulation results show that the proposed method can perform underdetermined DOA estimation with high accuracy.3.The covariance matrix of the array output can be transformed to a real-valued covariance matrix via unitary transformation.Discreting the spatial space,we can obtain the real-valued over-complete dictionary via unitary transformation.Then the real-valued covariance vector can be sparsely represented in the over-complete dictionary.The?₁norm optimization method is used to recovery the sparse vectors and estimate the DOA of the targets.The variant of the method can be applied to DOA estimation in non-uniform noise.Simulation results show the proposed method has good performance at low SNR and small snapshots,and the proposed method can estimate the DOA in non-uniform noise.4.To reduce the computation cost of the SBL?Sparse Bayesian Learning?based DOA estimation method,the DOA estimation method based on SBL using the real-valued covariance vector is proposed.The real-valued covariance vector can be sparsely represented in a over-complete dictionary when the received signals are uncorrelated,and the USMV CV-RVM?Unitary Single Measurement Vector Covariance Vector-based Relevance Vector Machine?method is proposed to estimate the DOA of uncorrelated signals.When the signals are coherent,the array covariance matrix is sparse represented,the sparse vectors of every column of the covariance matrix are jointly sparse recovered,the UMMV CV-RVM?Unitary Multiple Measurement Vectors Covariance Vector-based Relevance Vector Machine?method is proposed to estimate the DOA of coherent signals.The number of the targets that can be estimated using the proposed method is also analyzed.Simulation results show that USMV CV-RVM has a comparable performance with SMV CV-RVM with low complexity,and has better performance at low SNR and small snapshots.For correlated signals,the DOA estimation performance of UMMV CV-RVM is better than MMV CV-RVM,because the unitary transformation is helpful to decorrelate the coherent signals.5.The WSBL?Wideband Sparse Bayesian Learning?method is studied,the DOA estimation problem is constructed as a block sparse recovery problem,according to the property of the SBL method,the relative power spectrum estimation is updated with the iteration of the SBL method.The WSBL method is proposed to estimate the DOA and relative power simutaneously.The proposed method has good performance at low SNR and small snapshot number and can estimate the relative power at different frequencies efficiently.The computation cost of the proposed method is comparable with M?Multiple?SBL.Because of the estimated power accelerate the convergence speed,WSBL is faster than MSBL in simulation.