多域流形建模阵列普适性低复杂度DOA估计及非渐进性能界研究

This project is aimed at improving the robustness of super-resolution direction-of-arrival (DOA) estimators with array structure as well as reducing its complexity. Four novel manifold analytical methods in electrical-, slope-, logarithmic-, and real-valued- domains are proposed to reformulate the data model for DOA estimation, based on which the main contributions of the work are presented comprehensively and thoroughly. These includes five series of research contents focused on construction of DOA finding cost function with spatial aliasing and efficient search method for compressed spectral, fast two-dimensional (2-D) DOA estimation based on reduced one-dimensional (1-D) optimization, spectral factorization and selective polynomial rooting technique, blocked subspace decomposition on circulant array covariance matrix and low-cost DOA estimation under single I/Q framework, non-asymptotic theoretical performance bounds analysis on super-resolution DOA estimation algorithm. During the above researches, three scientific technologies are solved effectively, which includes transformed and fast estimation for signal/noise subspace, mathematical function finding for relating azimuth and elevation angles, ambiguity solving under single I/Q framework. The present work is able to provide new ideas for saving the hardware and software costs of real-world DOA estimation system, and is good for promoting its development. The present work is also of great significance for our country to improve the performance of military- and civilian- equipment such as radar, sonar, and communication system.

本项目以提高超分辨DOA估计算法对阵列结构的普适性和降低算法的复杂度为主要研究目的，在采用电气、斜率、对数、实值等多域流形建模方法改进阵列测向数据模型的基础上，全面、深入地开展模糊谱函数构造及谱峰压缩搜索、二维测向分批次一维化降维处理、谱分解及多项式选择性降阶求根、协方差阵循环分块及单通道测角探索、超分辨算法非渐进性能界理论分析等研究内容，同时着力突破子空间多域变换及快速估计、二维俯仰角和方位角函数关系确立、单通道测向解模糊等一系列科学问题，进而提出新型具有阵列结构普遍适应性的低复杂度算法，为超分辨DOA估计技术的工程化进度推进提供理论和应用参考。本项目研究成果有助于大幅节减测向系统的软硬件实现成本、提高其工作效率、推动其工程化进度。本研究对全面提高我国雷达、声呐等武器装备及移动通信等民用测向系统的实际工作性能同样具有十分重要的现实意义。

提升算法对阵列结构的适应性、增强算法对环境的稳健性以及降低算法的计算量，是超分辨阵列测向技术研究的核心任务，也是工程化推广过程中亟待解决的关键科学问题。通过四年的研究，我们提出了众多新颖的实用化超分辨测向算法，丰富了阵列测向的理论框架，不仅顺利完成了申请书中的既定目标，如：多域阵列流型解耦合、测向模型多域协同改进、模糊谱解模糊、多维参数分维处理、多项式选择性降阶求根、协方差矩阵降维分解等科学问题，同时，超额完成了单通道测角解模糊、酉变换完备实值处理、参数谱重构、子阵与整体阵列的子空间线性变换、实值与复值子空间的映射关系、特征提取多项式降阶和阵列流型可靠性近似等拓展研究内容，达到了预期的各项研究目标，并取得了丰富而显著的研究成果。截至目前，我们共培养博士生1人、硕士生6人，授权国家发明专利5项，受理专利3项，获省部级科研奖励2项，地市级科研奖励1项，参加学术会议1次，发表项目资助论文13篇，其中SCI论文12篇，会议论文1篇，SCI论文中IF>7的1篇，IF>4的3篇，IF>3文4篇。此外，我们将本项目研究成果积极推广，签约相关纵向科研项目4项，争取科研经费600余万元。