Sparse Bayesian Learning Theory And Its Application In Underwater Acoustic Array Signal Processing

In this paper,sparse reconstruction and underwater acoustic array signal processing are taken as the core,and the sparse reconstruction algorithm is studied with the aim of signal sparsity and high-precision direction of arrival(DOA)estimation.Based on the sparse Bayesian learning(SBL)theory,this paper studies the processing performance of underwater acoustic narrow-band sound pressure signals,underwater acoustic wide-band sound pressure signals and underwater acoustic broadband vector signals.Under the condition of low signal-to-noise ratio(SNR)and a small number of signal samples,the SBL algorithm outperforms the conventional beam-forming(CBF)algorithm and the existing compressed sensing sparse reconstruction algorithm when dealing with underwater acoustic signals.Therefore,the following studies are carried out in this paper.First,the SBL theory is introduced,then,the SBL algorithm and the l1 norm regularization method are introduced.The performance of each algorithm in terms of convergence,computational efficiency and angular estimation accuracy is investigated by simulation.The simulation results show that the SBL algorithm has better convergence performance than the M-SBL algorithm.The simulation results show that the SBL algorithm can estimate the DOA of the signal effectively compared to the CBF algorithm and the l1 norm regularization method at low SNR and small snapshot.Subsequently,the application of the algorithm in underwater acoustic array signal processing is analyzed,and the DOA estimation performance of the SBL algorithm and the l1 norm regularization method for the sound pressure signal and the vector signal is analyzed.The accuracy of vector signal estimation by the CBF algorithm and the SBL algorithm is compared by simulation experiments.The simulation analysis algorithm performs DOA estimation performance on narrow-band sound pressure signals,narrow-band coherent signals,wide-band sound pressure signals,and vector signals.The effects of SNR and snapshot on DOA estimation accuracy and computational efficiency of the CBF algorithm and sparse reconstruction algorithm are studied.The simulation results show that compared with the CBF algorithm and the l1 norm regularization method,the SBL algorithm can estimate the DOA of the signal effectively.Finally,the external field underwater test data and the pool sonar imaging test data under real underwater acoustic environment are processed to investigate the algorithm performance.The results of the wideband vector signal processing show that the SBL algorithm can estimate the direction of the incoming data of the test data with higher accuracy under the same number of snapshots.The results of real experimental data processing show that the SBL algorithm is more conducive to processing real-time signals.The sonar imaging data of the pool is processed.The results show that compared with the CBF algorithm,the SBL algorithm can clearly distinguish the iron image by imaging result,which can effectively distinguish the iron itself and the water surface and the pool wall scattering echo,which is estimated by the SBL algorithm.The result is sparse,and the position of the iron can be directly judged by the test processing result.