基于智能天线阵列的实用化高效超分辨测向技术研究

With a great prosperity and development in the last four decades, there are many critical problems to be solved in its real-world implementations for super-resolution direction-of-arrival (DOA) estimation techniques. Among, it is of great significance to reduce the computational complexity as well as to enhance the universality of DOA estimators. The present work focuses on practicability and high efficiency, which aims to propose a series of novel algorithms for one-dimensional and two-dimensional super-resolution DOA estimates. The manifolds of some typical arrays are first analyzed to reveal the coupling between DOA and array geometries, and three transformations in electrical domain, slope domain and frequency domain are proposed to improve the data model of DOA estimate. With these improved models and our previous works, two new methods based on dimension reduction and spectral compression, respectively, are developed for fast DOA estimation with arbitrary array configurations. Moreover, new schemes for low-complexity subspace decomposition with real-valued computations and fast two-dimensional DOA estimation with one dimensional search techniques are discussed. The statistical performance of proposed approaches are finally theoretically analyzed, and Cramér-Lower bound (CRLB) and Mean Square Error (MSE) expression for DOA estimation are derived. The research in the present work not only provides new meanings for super-resolution theory, but also offers significant contributions for the implementation of direction estimation.

超分辨阵列测向技术虽然已经过近40年的蓬勃发展，但实践证明其工程化进程中尚有大量关键技术亟待攻克。其中，降低算法的计算量、提高算法的普适性是具有重要意义的研究工作之一。本项目以实用化为向导，拟对适用任意阵形和信号形式的高效超分辨阵列测向技术展开研究。首先，从典型阵形入手，分析导向矢量角度信息和阵列几何结构的耦合关系，结合电气、斜率和频率等变换域分析方法改进阵列测向的数据模型。其次，研究任意阵形下的矢量数据降维和空间谱搜索范围压缩方法，在此基础上论证以协方差矩阵分裂分解进行实值子空间估计的机理，探索二维优化问题的分批次一维化处理技术，以若干次一维测向实现二维阵列联合测向。最后，根据子空间统计特性，分析新算法的理论性能，推导其克拉美-罗下界和估计误差解析表达式。本项目研究结果将进一步丰富超分辨的理论内涵，同时将推进超分辨阵列测向技术的工程化应用进度。

降低算法计算量、提高算法普适性，是阵列信号处理领域具有重要研究意义的研究工作，也是推进超分辨阵列测向技术工程化进度的关键技术。通过三年的研究，我们提出了多种高效阵列测向新算法、新理论，不但顺利解决了申请书所规划的导向矢量构成方式分析、测向数据模型改进、矢量数据降维、空间谱搜索范围压缩、矩阵分裂分解和二维测向分批次一维处理等科学问题，而且超额完成了酉变换改进、子空间旋转变换、多项式降阶、实值化求根、快速子空间估计和子空间交集求解等拓展研究内容，达到了预期研究目标，取得了丰富而显著的研究成果。截至目前，我们参加各类学术会议3次，培养博士生和硕士生各2人，发表和录用项目资助论文24篇。发表的论文中，SCI检索18篇，EI检索3篇，会议3篇，SCI论文IF>3的论文4篇，IF>2的论文3篇，SCI论文web of science总计引用33次。此外，在本项目研究基础上，我们将项目研究成果积极推广和转化，签约相关科研项目3项，争取科研经费260余万元。