节点文献
多阵列分布源参数估计及跟踪方法研究
Research of Parameter Estimation and Tracking Method for Distributed Source Using Multiple Arrays
【作者】 郭贤生;
【导师】 万群;
【作者基本信息】 电子科技大学 , 信息获取与探测技术, 2008, 博士
【摘要】 分布源现象广泛存在于移动通信、雷达、声纳等领域。近年来众多学者提出了多种分布源模型及参数估计方法,但其中的大部分算法都存在以下问题:首先这些算法大都针对一维分布源的参数估计问题,并且有的仅考虑中心波达方向(DOA)估计,没有考虑角扩展参数估计问题;其次,同时估计中心DOA和角扩展参数的算法又需要一维或二维搜索,计算复杂度高,不易实现。如何降低分布源参数估计算法的复杂度一直是本领域的研究热点。针对上述问题,本文在分析多阵列观测的分布源模型基础上,提出几种低复杂度的分布源参数估计方法。为了获取分布源中心DOA发生变化时的角度估计,本文还给出了一种快速的分布源中心DOA跟踪方法。本文的主要创新点概括如下:1.提出了一种低复杂度一维相干分布源参数估计方法。用双均匀线阵做接收阵列,首先推导了双线阵关于分布源中心DOA的近似旋转不变性,利用ESPRIT(Estimating Signal Parameter via Rotational Invariance Techniques)类方法得到分布源的中心DOA估计,然后用FOCUSS(Focal Underdetermined System Solver)方法得到分布源的角扩展参数估计。由于无需搜索,该方法与基于二维搜索的DSPE(Distributed Signal Parameter Estimator)方法相比,复杂度显著降低。另外,还把分布源的中心DOA估计问题转化为近似稀疏表示问题,利用FOCUSS方法得到中心DOA的快速估计。2.提出了一种低复杂度二维相干分布源DOA解耦估计方法。用双平行均匀线阵做接收阵列,利用分布源广义方向矢量的二次旋转不变性(Quadric RotationalInvariance Property,QRIP)和其在中心DOA上的一阶泰勒级数展开得到的近似旋转不变性,提出了一种能同时估计中心俯仰角和中心方位角的解耦DOA估计方法,该算法不需搜索,适用于分布源具有不同角信号分布情况或角信号分布未知情况。另外,还提出了一种基于极小最小特征值和极大最大特征值的参数配对方法。3.提出了一种低复杂度二维非相干分布源参数估计方法。用双平行均匀圆阵做接收阵列,首先推导了双均匀圆阵在中心俯仰角上的近似旋转不变性,在此基础上提出用于估计中心俯仰角的修正TLS-ESPRIT方法:其次,构造一种新的一维广义MUSIC谱,并通过搜索得到中心方位角估计;最后,俯仰角扩展和方位角扩展通过协方差矩阵匹配得到。该方法估计四维参数只需一次一维搜索,能够估计多个具有不同角功率密度函数的分布源参数,且复杂度低、不存在参数配问题。4.针对分布源DOA发生变化时的跟踪问题,提出了一种基于子空间更新的分布源中心DOA的跟踪方法。仿真结果表明,该方法在分布源的角扩展较小时能够给出较精确的跟踪结果。
【Abstract】 Distributed source phenomenon always appears in radar, sonar and wireless communication fields. Several distributed models and many parameter estimation algorithms have been proposed in recently years. However, most of them have some problems as follows: Firstly, many algorithms are designed for one-dimensional (1D) distributed sources and some of them estimate the central direction-of-arrivals (DOAs) only. Secondly, the algorithms that can estimate both the central DOAs and angular spreads all need 1D or 2D search, which will bring heavy computation burden. How to build a low-complexity algorithm is the research hot topic in distributed source field.Based on the previous work, this dissertation proposes several new low-complexity algorithms for distributed source using multiple arrays. In addition, a fast DOAs tracking algorithm is also addressed in the last part. The main content is summarized as follows:1. A low-complexity parameter estimation method is proposed for 1D coherently distributed (CD) sources. Based on two uniform linear arrays (ULAs), we derive an approximate rotational invariance property with respect to the central DOAs, which can be used to obtain the central DOAs estimation with ESPRIT (Estimating Signal Parameter via Rotational Invariance Techniques) type algorithms. The angular spreads are estimated by FOCUSS (Focal Underdetermined System Solver). Because there is no search operator, the computational complexity of the proposed method is much lower than that of DSPE (Distributed Signal Parameter Estimator) algorithm. In addition, we also cast the central DOAs estimation problem into sparse signal representation frame to obtain the central DOAs estimation.2. A low-complexity 2D CD sources decoupled DOAs estimation approach is proposed. Based on two parallel ULAs, we use the approximate rotational invariance property, which is derived from the one-order Taylor series expansion of generalized steering vector at the central DOAs, and the quadric rotational invariance property (QRIP) of generalized steering vector to realize a decoupled DOAs estimation. The proposed method needs not search, and it is suitable for unknown angular signal distribution case. In addition, we also present two parameters matching methods: min-minimum eigenvalue and max-maximum eigenvalue method to solve the DOAs matching problem.3. A low-complexity parameter estimation method for 2D ID sources is proposed. Based on two parallel uniform circular arrays (UCAs), we derive an approximate rotational invariance property with respect to the central elevation DOAs, with which the central elevation DOAs of ID source can be estimated using TLS-ESPRIT, the central azimuth DOAs is estimated by constructing a novel 1D generalized MUSIC (GMUSIC) spectrum. Based on the preliminary estimation results, the angular spreads can be given by covariance matrix matching technique. Our method is suitable for unknown angular signal distribution case. In addition, the complexity of our method is low and there is no parameters matching problem.4. To obtain the central DOAs estimation when the location of distributed source varies, we also proposed a fast central DOAs tracking method for CD and ID source based on subspace updating. Simulation results show that our method can give more exact DOAs tracking results for small angular spread.
【Key words】 distributed source; central DOA; angular spread; subspace updating; sparse component analysis (SCA);