基于压缩感知的水声信号波达方向估计方法研究

发布时间：2018-08-19 10:40

【摘要】：波达方向(DOA)估计作为阵列信号处理的重要研究内容之一,在水声信号处理方面有着广阔的应用前景,突破了传统奈奎斯特采样频率约束的压缩感知(CS)理论为信号处理领域注入了新思想,基于CS理论的DOA估计已成为一个新的研究方向,其对水声信号的探测、定位和跟踪具有重要的理论和现实意义。在此背景下,本文基于CS理论框架开展了水声信号DOA估计的相关研究工作,主要研究成果有:针对检验本文理论成果价值和实际应用场景的需求,在研究了舰船辐射噪声产生的原因和频谱特点分析的基础上,建立了相应的数学模型并模拟出舰船辐射噪声的时域波形和频谱,以此作为本文理论应用的水声信号对象。针对传统阵列信号处理的MMUSIC和ES-DOA算法硬件处理系统繁多运算量较大带来的测量成本较高的问题,提出了基于CS框架下的DOA估计算法。利用测量矩阵采样阵列输出的同时对数据进行压缩,由此得到了一种可压缩阵列(CSA)结构,该结构可以减少前置电路的数量,降低测量的硬件成本。在此CSA结构基础上,提出了可压缩阵列的MMUSIC(CSA-MMUSIC)和可压缩阵列的ES-DOA(CSA-ES-DOA)两种DOA估计算法,理论分析表明这两种算法均可以降低软件运算量,仿真实验验证了这两种算法可使用较少的采样数据获得与原算法相近的性能,并且可在低快拍、低信噪比和小测量矩阵的行数(M)下,较好的实现对相干和邻近水声信号的DOA估计。针对随机测量矩阵所需存储空间大且硬件实现困难的问题,提出了一种基于光正交码(OOC)的构建确定性矩阵的方法。首先将OOC和Singer结构结合建立基矩阵,再用Bernoulli矩阵扩展基矩阵得到确定性矩阵OOC-B,证明了该矩阵在一定条件下具有渐近最优性,分析了其列相关性,并给出了特定参数的矩阵实例。其次,为了研究OOC-B作为确定性测量矩阵的性能,将其用于CSA-MUSIC的水声信号DOA估计中,仿真结果表明,相对于OOC-H和随机高斯矩阵,使用OOC-B作为测量矩阵具有更优的估计性能,且在低信噪比和小快拍下具有较小的估计偏差。最后,将OOC-B作为测量矩阵用于空域压缩重建的水声信号DOA估计,仿真结果表明在相同信噪比下使用OOC-B的DOA估计偏差更小且成功恢复率更高。针对声压传感器线型阵列MMUSIC算法的左右舷模糊的问题,提出了基于CS理论的矢量水听器阵列DOA估计算法。利用确定性测量矩阵OOC-B采样矢量水听器的输出,构建了确定性压缩采样矢量水听器阵列(DCV)结构,该结构可以降低电路成本,在此结构基础上,提出了确定性压缩采样的矢量水听器阵列MMUSIC(DCV-MMUSIC)算法并进行仿真验证,结果表明该算法具有运算量较低和抑制左右舷模糊的优点,可以在低信噪比、小快拍和小M下较好地实现相干水声信号的DOA估计。并且,使用该算法对实际商船辐射噪声信号进行了DOA估计,仿真结果表明在不同信噪比、快拍数和M下商船辐射噪声的估计偏差虽略大于模拟水声信号的,但仍然具有较好的估计性能。
[Abstract]:Direction of arrival (DOA) estimation, as one of the important research contents of array signal processing, has broad application prospects in underwater acoustic signal processing. It breaks through the traditional compression sensing (CS) theory with Nyquist sampling frequency constraints and injects new ideas into the field of signal processing. DOA estimation based on CS theory has become a new research direction. Under this background, based on the CS theory framework, this paper carries out the related research work of underwater acoustic signal DOA estimation. The main research results are as follows: According to the requirements of testing the theoretical value and practical application scenarios, the ship radiated noise production is studied. Based on the analysis of the causes and spectrum characteristics, a mathematical model is established to simulate the time-domain waveform and spectrum of ship radiated noise, which is used as the underwater acoustic signal object in this paper. In this paper, a DOA estimation algorithm based on CS framework is proposed. A compressible array (CSA) structure is proposed by using the measurement matrix to sample the array output and compress the data at the same time. This structure can reduce the number of pre-circuits and reduce the hardware cost of measurement. MMUSIC (CSA-MMUSIC) and ES-DOA (CSA-ES-DOA) estimation algorithms for compressible arrays are proposed. Theoretical analysis shows that both algorithms can reduce software computation. Simulation results show that the two algorithms can achieve similar performance with the original algorithm by using fewer sampling data, and can also achieve low snapshot, low signal-to-noise ratio and small measurement matrix. In order to estimate the DOA of coherent and adjacent underwater acoustic signals better under the number of rows (M), a method of constructing deterministic matrix based on optical orthogonal code (OOC) is proposed to solve the problem of large storage space and difficult hardware implementation of random measurement matrix. The deterministic matrix OOC-B is obtained, and the asymptotic optimality of the matrix is proved under certain conditions. The column correlation of the matrix is analyzed, and an example of the matrix with specific parameters is given. Secondly, in order to study the performance of OOC-B as deterministic measurement matrix, it is applied to DOA estimation of underwater acoustic signals of CSA-MUSIC. The simulation results show that the matrix has asymptotic optimality relative to OOC-H. And the stochastic Gaussian matrix, using OOC-B as the measurement matrix has better estimation performance, and has smaller estimation bias under low SNR and small snapshot. Finally, using OOC-B as the measurement matrix for underwater acoustic signal DOA estimation, simulation results show that the DOA estimation error using OOC-B under the same SNR is smaller and the estimation error is smaller. A DOA estimation algorithm for vector hydrophone array based on CS theory is proposed to solve the problem of port-starboard ambiguity in MMUSIC algorithm of linear array of sound pressure sensors. A deterministic compressed sampling vector hydrophone array (DCV) structure is constructed by using the output of OOC-B sampling vector hydrophone with deterministic measurement matrix. On the basis of this structure, a deterministic compression sampling vector hydrophone array MMUSIC (DCV-MMUSIC) algorithm is proposed and simulated. The results show that the algorithm has the advantages of low computational complexity and suppression of port-starboard ambiguity. The DOA of coherent underwater acoustic signals can be realized well in low signal-to-noise ratio, small snapshot and small M. Furthermore, DOA estimation of the actual merchant ship radiated noise signal is carried out by using this algorithm. Simulation results show that the estimation error of merchant ship radiated noise under different SNR, snapshot number and M is slightly larger than that of the analog underwater acoustic signal, but still has better estimation performance.
【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN911.7

【相似文献】