Research On The Algorithm Of Wideband DOA Estimation Via Sparse Bayesian Inference

Direction-of-arrivals estimation is a key part of underwater acoustic array signal processing,and it is an important part of realizing accurate underwater target detection.Traditional DOA estimation methods have poor estimation performance under the conditions of low signal-to-noise ratio,small snapshots,and correlated signals.The DOA estimation algorithm based on sparse Bayesian learning that has appeared in recent years has solved these problems well.However,most of the current improvements are based on research on narrowband signal detection.This paper expands on this basis,and launches the research of wideband signal DOA estimation algorithm based on sparse Bayesian learning.In order to reasonably extend the existing algorithms of narrowband signals based on sparse Bayesian learning to the field of wideband signal processing,this paper derives the narrowband signal SBL algorithm working in the frequency domain(SBL-FD),and verifies the DOA estimation algorithm based on SBL through simulation.And the SBL-FD algorithm also has good estimation performance in the case of low signal-to-noise ratio and small snapshots,and can process coherent signals.Next,this paper introduces the concept of spatial sparseness into the frequency division model of wideband signals,establishes the M~4 model of wideband signal sparse reconstruction,and derives the SBL-based wideband signal DOA algorithm(w SBL).Aiming at solving the common off-grid problem in the sparse reconstruction DOA algorithm,that is the problem that the signal incident angle does not match the spatial gridding,the study introduces the first-order Taylor expansion model and linear interpolation model into the broadband signal sparse reconstruction model,and derives the formulas for estimating the off-grid error under this two method.In addition,considering the problem of unknown frequency bands in the scene of detecting broadband non-cooperative signals,an energy threshold weighting(ETW)signal preprocessing method is proposed.The simulation results show that the two off-grid models can greatly reduce the estimation error caused by the sparse spatial grid.On this basis,the ETW preprocessing is added to improve the iterative convergence speed and shorten the running time.A key problem of using the frequency division model of the wideband array signal to estimate the DOA is that the amount of calculation is too large.On the basis of the above research,this paper gives some solutions in the array element domain and the covariance matrix domain.When processing array element domain data,SVD decomposition or unitary transformation real-valued method is used to further process the signal after the ETW method.The former can directly reduce the amount of algorithm data under the premise of the known number of sources;the latter solves the problem of too slow operation speed caused by complex numbers,which actually increases the amount of algorithm data.In view of the still existing problems of the improved efficient DOA algorithm in the array element domain,this paper finally proposes a wideband signal SBL algorithm in the covariance matrix domain(w PCM),and uses the Toplitz property of the ULA covariance matrix to further reduce the amount of calculation.In order to solve the problem that the Toplitz property is only applicable to independent signal scenarios,the preprocessing method of understanding coherence is added to solve this problem.Simulation experiments have proved that the above algorithms greatly shorten the running time of the algorithm on the basis of almost unchanged the estimation accuracy of the algorithm in different scenarios.Finally,the feasibility of the above algorithm is verified by processing sea trial data.