基于空间稀疏的匹配场源定位

发布时间：2018-12-07 17:54

【摘要】：匹配场处理是目前水下被动源定位的主要手段之一。常规匹配场处理方法分辨能力较低、旁瓣较高,因此提高分辨力的问题作为匹配场处理技术的一个重要目标,受到各国研究者的广泛关注。压缩感知理论提出了一种全新的信息获取指导理论,打破了传统采样定理的限制,用随机采样获取信号的样本,然后通过非线性重建算法完美的重建信号。该理论进一步完善和丰富了信号稀疏表示理论,是现代信息理论的一个重大变革。因此本文结合稀疏重构理论研究成果,开展了基于空间稀疏的匹配场声源定位方法研究。本论文首先基于匹配场定位搜索空间的稀疏性,结合海洋声场模型,开展了基于空间稀疏理论的匹配场定位信号的模型研究。其次由于稀疏重构理论多采用基于l1范数的稀疏表示模型,采用该模型目的是为了求解方便,而该模型计算量较大,运算时间较长,因此本文又研究了基于l0范数的匹配场源定位方法,并且针对l0范数方法计算复杂度高的问题提出一种快速的算法,即平滑l0范数匹配场源定位方法,通过仿真实验验证了该方法可行性和有效性,并且证明该方法重构精度高,而且具有较高的空间分辨率,在低信噪比下仍具有较好的估计性能,弥补了现有基于空间稀疏的匹配场源定位方法中运算速度慢、算法复杂度高等问题。然后,针对平滑l0算法近似l0范数的估计存在不精确、收敛速度慢的问题,在研究平滑l0范数原理的基础上,对其做出了改进,选取更具陡峭性的逼近函数,提高收敛速度,并且结合阻尼牛顿法,提出了一种基于改进平滑l0范数的匹配场源定位方法。通过仿真实验,说明该方法能够大幅度减少稀疏重构信号的运算的复杂度,减少了定位所用的时间。相比于平滑l0范数,改进的平滑l0范数定位的成功率更高。最后,针对基于空间稀疏的匹配场源定位方法主要只利用了空间稀疏性,并没有考虑到信号自身结构的稀疏特性。当对声源进行空间定位时,目标信号自身会呈现出块结构的稀疏特性。因此本论文的另一方面工作就是基于多观测模型,建立匹配场声源定位的结构稀疏表示模型,提出一种基于结构稀疏的匹配场源定位方法。通过仿真实验表明该方法重构精度高具有较高的空间分辨率,在低信噪比下仍然具有较好的定位性能。
[Abstract]:Matching field processing is one of the main methods of underwater passive source location. The resolution of conventional matching field processing methods is low and the sidelobe is high. Therefore, the problem of improving the resolution is an important target of the matching field processing technology, which has been widely concerned by researchers all over the world. Compression perception theory proposes a new information acquisition guidance theory, which breaks the limitation of the traditional sampling theorem, uses random sampling to obtain samples, and then uses nonlinear reconstruction algorithm to reconstruct the signal perfectly. This theory further improves and enriches the signal sparse representation theory and is an important change in modern information theory. Therefore, based on the research results of sparse reconstruction theory, this paper studies the location method of matched field sound source based on spatial sparsity. Firstly, based on the sparsity of the matching field location search space and the ocean sound field model, the model of the matching field location signal based on the spatial sparsity theory is studied in this paper. Secondly, the sparse representation model based on L1 norm is used in sparse reconstruction theory. The purpose of this model is to solve the problem conveniently. Therefore, this paper also studies the matching field source location method based on l0 norm, and proposes a fast algorithm, that is, smooth l0-norm matched field source localization method, aiming at the problem of high computational complexity of l0-norm method. The simulation results show that the proposed method is feasible and effective, and the reconstruction accuracy is high, the spatial resolution is high, and the estimation performance is good at low signal-to-noise ratio (SNR). It makes up for the slow operation speed and high complexity of the existing matching field source location methods based on spatial sparsity. Then, aiming at the problem that the estimation of approximate l0 norm of smoothing l0 algorithm is imprecise and the convergence rate is slow, on the basis of studying the principle of smoothing l0 norm, an improvement is made on it, and a steeper approximation function is selected to improve the convergence speed. Combined with damped Newton method, a matching field source location method based on improved smoothing l0 norm is proposed. The simulation results show that the proposed method can greatly reduce the computational complexity of sparse reconstruction signal and reduce the time of location. Compared with smooth l0 norm, the success rate of improved smooth l0 norm localization is higher. Finally, the spatial sparseness of the matched field source location method based on spatial sparsity is mainly used, and the sparse property of the signal structure is not considered. When the sound source is located in space, the target signal itself presents the sparse property of block structure. Therefore, the other work of this paper is to establish a structural sparse representation model of matched field sound source location based on multi-observation model, and to propose a method of matching field source location based on structural sparsity. The simulation results show that the reconstruction method has higher spatial resolution and better location performance under low SNR.
【学位授予单位】：江苏科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TB56

【相似文献】