Acoustic Vector Sensor Array Signal Processing Based On Compressive Sensing

Acoustic vector sensor can simultaneously measure the acoustic pressure information and the acoustic particle velocity information at a point in the acoustic field. Compared with the traditional pressure sensor, acoustic vector sensor can get more information from the acoustic field and effectively improve the performance of underwater acoustic systems. Acoustic vector sensor and its array signal processing technology has been widely applied in the field of underwater acoustic engineering. The compressive sensing (CS) theory raised in recent years, is a new theory of sparse signal compression and restoration, which breaks through the limitations of traditional nyquist sampling theorem. The CS theory requires that the data is sampled and compressed at the same time, which reduces the requirements of the sampling system. The CS theory has become a new research hotspot in the field of signal processing. Therefore, the acoustic vector sensor array and CS theory are applied to the underwater source localization model to get better positioning performance. It is necessary to research the underwater source localization algorithms based on CS by using the acoustic vector sensor array. The main research contents of this paper are as follows:Firstly, we introduce the main contents of the CS theory and the acoustic vector sensor far-field array-manifold and near-field array-manifold in detail. Moreover, we briefly introduce the matrix algebra knowledge involved in the source localization algorithms presented in this paper, signal model and the characteristics of underwater acoustic channel.Then, we begin to research the far-field source DOA estimation algorithms based on acoustic vector sensor array. The far-field acoustic vector sensor array signal processing mathematical model is given firstly. We research on three traditional DOA estimation algorithms based on acoustic vector sensor array:CBF algorithm, MVDR algorithm and MUSIC algorithm. Then, we do some experimental simulations to analyze the DOA estimation performance of the three algorithms. According to the sparsity of underwater far-field sources azimuth in the whole angle scanning space, the CS theory is applied to the acoustic vector sensor array DOA estimation model. The corresponding model is established. The DOA estimation of the underwater sources is realized by using the l1-SVD algorithm. Simulation experimental results show that the CS algorithm can achieve DOA estimation with high resolution for sources in the case of small snapshots, low signal to noise ratio (SNR) and signal coherence, and accurately resolve closely spaced signals, better than the traditional DOA estimation algorithms. The results verify the feasibility and effectiveness of the CS algorithm.Finally, we research the near-field source localization algorithms based on acoustic vector sensor array. The near-field acoustic vector sensor array signal processing mathematical model is given firstly. We research the near-field source localization algorithms based on two-dimensional search, and analyze the positioning performance of the two-dimensional MVDR near-field source localization algorithm and the two-dimensional MUSIC near-field source localization algorithm by using the simulation results. Considering the high computational complexity of the two-dimensional search algorithms, we propose the MUSIC near-field localization algorithm based on second-order statistics. In the proposed algorithm, by exploiting the second-order statistics of the acoustic vector sensor array outputs, we construct a far-field approximate covariance matrix in relation with the source azimuth only. By separating the azimuth parameter and range parameter of sources, we estimate the azimuth of near-field sources firstly. Then, we use the azimuth estimation to estimate the range of near-field sources. In our proposed algorithm, the previous two-dimensional search positioning problem is converted to multiple one-dimensional search estimation problems, which significantly reduces the computational complexity. According to the sparsity of near-field sources in the whole near-field region, the CS theory is applied in the acoustic vector sensor array near-field source localization model. Then, we establish the acoustic vector sensor array near-field localization model based on CS. In order to reduce the computation, the model is improved by using the thought of separating the azimuth parameter and range parameter of sources. The azimuth estimation and range estimation are obtained respectively by using the i1-SVD algorithm. Simulation results show that the CS algorithm can improve the estimation accuracy in the case of low SNR, outperforming the MUSIC near-field localization algorithm based on second-order statistics. The results verify the validity of the CS algorithm.