Applications Of Low Complexity Compressive Sensing Technology In Underwater Acoustic Array Signal Processing

Direction of arrival(DOA)estimation is one of the important means of underwater acoustic array signal processing.It is widely used in underwater target detection,signal parameter estimation,sonar imaging and other fields.In view of the sparse characteristics of DOA in underwater acoustic array signal processing,this paper applies compressed sensing sparse reconstruction algorithms to DOA estimation and sonar imaging fields.Meanwhile,research and improve the algorithms.First of all,the paper introduces the main content of compressed sensing theory and several greedy iterative algorithms,such as the orthogonal matching pursuit(OMP),the orthogonal matching pursuit using dichotomous coordinate descent(OMP-DCD),the multipath matching pursuit(MMP),and the MMP-DCD algorithm.Analyze the sparse reconstruction performance of each algorithm.Then,in the DOA estimation,the performances of the Capon algorithm and the compressed sensing algorithms for small snapshots and coherent sources are analyzed.For the problem of the estimated performance degradation caused by the off-grid,using the dictionary learning(DL)method,the OMP-DCD-DL algorithm is given based on the acoustic pressure sensor array.The research shows that this algorithm can effectively improve the performance of the compressed sensing DOA estimation under the off-grid condition by updating dictionary iteratively.In sonar imaging,the Back projection(BP)algorithm is introduced.Under the distributed multiple input multiple output(MIMO)sonar model,the imaging performance of the BP algorithm and the OMP-DCD algorithm is analyzed.In order to solve the problem of the computational complexity caused by the small mesh size of the sonar imaging area,using the space alternating(SA)method,the SA-OMP-DCD algorithm is given.The research shows that the algorithm can locate the targets with high precision while reducing the amount of calculation.Finally,the experimental data processing is carried out.The experimental results of broadband vector signal show that,compared with the Capon algorithm,compressed sensing algorithms can estimate the DOA more accurately with single-snapshot of one frequency point,which both saves the storage and transmission space,and more conducive to real-time signal processing.The results of near-field targets imaging sonar processing show that,the OMP-DCD,MMP-DCD and OMP-DCD-DL algorithms can distinguish two close targets with high resolution,and has higher resolution than FFT algorithm and Capon algorithm.