阵列测向的稀疏超分辨方法研究

发布时间：2018-09-12 05:51

【摘要】：在现代复杂的空间电磁环境中,辐射源信号高度密集、目标高速运动、“短,跳,隐”信号大量出现,作为电子侦察与电子对抗重要组成部分的无线电测向定位遇到了前所未有的困难与挑战。本文研究完善并利用压缩感知理论,针对低信噪比、少快拍数、目标运动等问题,构建测向模型,设计稀疏重构算法,提出了稀疏超分辨的阵列测向与方位角跟踪的新方法。首先,基于?-约束等距性质研究了?_1-analysis稀疏重构的可重构条件。针对目前?_1-analysis稀疏重构所要满足的?-约束等距性质本质上要求测量矩阵具有相关性很小这一限制,本文给出了?-约束等距约束常数的紧致上界,松弛了测量矩阵的相关性条件,为?_1-analysis稀疏重构应用于实际问题提供了理论保障。其次,提出了低信噪比环境下阵列接收信号的降噪恢复方法。在信噪比很低的情况下,传统的空间谱估计方法和基于压缩感知理论的稀疏测向方法的抗噪性能均不理想。本文建立了适用于阵列接收信号恢复的?_1-anaylsis重构模型,证明了阵列接收模型中的流形矩阵满足?_1-anaylsis稀疏重构条件,从而在理论上保证了将?_1-analysis稀疏重构用于低信噪比环境下恢复阵列接收信号的合理性,最后推导出了信号恢复误差的理论上界。实验表明该方法对提高低信噪比环境下的测向性能具有显著效果。第三,分别提出了基于信号子空间与信号结构信息的空间离散化阵列稀疏测向方法。目前大多数稀疏测向方法将阵列接收数据或其协方差矩阵作为稀疏重构模型中的观测数据,测向的准确性不理想,针对这一问题本文提出了利用信号子空间的阵列稀疏测向新方法,构造了信号子空间的稀疏表示,建立了恢复辐射源信号能量的稀疏重构模型,并将其转换为二阶锥规划问题来求解。为了利用辐射源信号包括空间稀疏性和时间相关性在内的信号结构信息,本文提出了适用于阵列测向的块稀疏贝叶斯方法,该方法同时结合了空间网格细化策略,在一定程度上缓解了大部分阵列稀疏测向方法所面临的由空间离散化导致的辐射源目标网格失配和稀疏表示模型误差这一问题,进一步提高了测向的性能。再次,提出了连续域阵列测向的稀疏超分辨方法。基于空间离散化的稀疏测向方法的性能受到网格失配效应和稀疏表示模型误差的限制,为了解决这一问题,本文基于无网格压缩感知理论,利用单快拍数据实现了连续域稀疏超分辨测向,并给出了该方法实现超分辨测向所要满足的辐射源最小角度距离和最少阵元个数。针对上述方法的测向性能受到相邻辐射源之间最小角度距离和阵元数限制且仅能使用单个快拍数据使得性能受限等问题,本文提出了可以使用任意快拍数据实现连续空间上阵列测向的稀疏超分辨方法。最后,提出了基于压缩稳健主成分分析的运动目标波达角跟踪方法。目前鲜见专门针对运动目标的波达角跟踪的研究,本文利用压缩稳健主成分分析理论将波达角跟踪问题可以转化为一个低秩矩阵和稀疏矩阵恢复的问题,并设计了数值求解该优化问题的线性交替方向法,实现了从阵列接收数据中恢复出固定目标信号矩阵和运动目标信号矩阵,并根据恢复的信号矩阵进一步确定固定目标的波达角估计、跟踪运动目标的波达方向。
[Abstract]:In the modern complex space electromagnetic environment, the radiator signal is highly dense, the target is moving at high speed, "short, hop, hidden" signals appear in large numbers, as an important part of electronic reconnaissance and electronic countermeasures, radio direction finding and positioning has encountered unprecedented difficulties and challenges. In this paper, a direction finding model is constructed and a sparse reconstruction algorithm is designed. A new method of sparse super-resolution array direction finding and azimuth tracking is proposed. Firstly, the reconfigurable condition of?_1-analysis sparse reconstruction is studied based on?-constraint equidistant property. The isometric property essentially requires that the measurement matrix has very little correlation. In this paper, a compact upper bound of the isometric constraint constant is given, which relaxes the correlation condition of the measurement matrix, and provides a theoretical guarantee for the application of the?_1-analysis sparse reconstruction to practical problems. Secondly, the array received signals in low SNR environment are proposed. In the case of low signal-to-noise ratio, the traditional spatial spectrum estimation method and sparse direction finding method based on compressed sensing theory have poor anti-noise performance. In this paper, a?_1-anaylsis reconstruction model for array received signal recovery is established, which proves that the manifold matrix in the array received model satisfies?_1-anaylsis sparse. The sparse reconstruction condition guarantees the rationality of applying?_1-analysis sparse reconstruction to recover the received signal of array in low SNR environment theoretically. Finally, the theoretical upper bound of the signal recovery error is deduced. Experiments show that this method has significant effect on improving the direction finding performance in low SNR environment. The sparse array direction finding method based on signal subspace and signal structure information is a sparse array direction finding method. Most of the sparse direction finding methods take the received array data or its covariance matrix as the observation data in the sparse reconstruction model. The accuracy of direction finding is not ideal. To solve this problem, this paper proposes the sparse array direction finding method based on signal subspace. The sparse representation of the signal subspace is constructed, and the sparse reconstruction model for recovering the energy of the emitter signal is established and converted into a second-order cone programming problem. Sparse Bayesian method, which combines spatial mesh refinement strategy, alleviates to a certain extent the problem of grid mismatch and sparse representation model error caused by spatial discretization in most sparse direction finding methods, and further improves the performance of direction finding. Thirdly, a continuous-domain array is proposed. Sparse super-resolution method for direction finding. Sparse direction finding method based on spatial discretization is limited by grid mismatch effect and sparse representation model error. To solve this problem, this paper uses single snapshot data to realize sparse super-resolution direction finding in continuous domain based on Meshless compressed sensing theory, and gives the method to realize super-resolution direction finding. In order to solve the problem that the minimum angle distance and the minimum number of array elements are required to satisfy the resolution of direction finding, which is limited by the minimum angle distance and the number of array elements between adjacent sources and can only use a single snapshot data, this paper proposes that the performance of the method is limited by using arbitrary snapshot data to achieve continuous space. Sparse super-resolution method for array direction finding. Finally, a method of DOA tracking for moving objects based on compressed robust principal component analysis is proposed. At present, there is little research on DOA tracking for moving objects. In this paper, the problem of DOA tracking can be transformed into a low rank matrix and sparse by using compressed robust principal component analysis theory. The linear alternating direction method is designed to solve the problem of matrix recovery. The fixed target signal matrix and the moving target signal matrix are recovered from the array receiving data, and the DOA estimation of the fixed target is further determined according to the recovered signal matrix to track the direction of arrival of the moving target.
【学位授予单位】：国防科学技术大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN911.7

【相似文献】