【作者】 郑桂妹；

【导师】 陈伯孝；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2014， 博士

【摘要】 电磁波的极化特征是信号幅度、相位、频率和波形等信息外,另一个可资利用的重要特征信息。对极化信息的充分挖掘和利用,有利于提高雷达和通信等系统的性能。电磁矢量传感器构成的极化敏感阵列在雷达、通信、声纳和生物医学等众多领域具有广阔的应用前景。另一方面,MIMO雷达以其在目标检测、参数估计等方面的独特优势,使其成为近十年的一个研究热点。而参数估计(如波达方向、极化状态角等)是雷达和通信等系统的主要任务之一。故本文主要研究基于电磁矢量传感器以及电磁矢量传感器在MIMO雷达应用中的波达方向、极化状态角等参数估计问题。具体内容可概括为如下四个部分。第一部分,研究电磁矢量传感器对相干信源波达方向的估计问题。首先建立传统共点式电磁矢量传感器对完全极化入射信源响应的信号模型,接着回顾几种基于共点式电磁矢量传感器的波达方向估计算法。然后针对传统极化平滑算法解相干源时没有利用子阵互相关信息导致分辨率较差的问题,提出一种新的解相干源预处理方法：加权极化平滑算法。该算法利用了电磁矢量传感器阵列的六个分量组成子阵的全部自相关和互相关信息,对接收阵列协方差矩阵的子矩阵做加权滑动平均,得到等效的阵列协方差矩阵,以该协方差矩阵对角化为约束,推导最优加权系数的理论表达式,并分析等效信源协方差矩阵的秩,得到加权极化平滑算法最大的解相干源数为6。计算机仿真结果表明加权极化平滑算法比传统的极化平滑算法具有更高的分辨性能和估计精度。第二部分,提出四种分离式电磁矢量传感器阵列结构,并研究其波达方向和(或)极化状态角联合估计问题。传统共点式电磁矢量传感器阵元间互耦效应明显,导致电磁矢量传感器硬件实现困难、波达方向和极化状态角估计性能严重下降。较之传统的共点式电磁矢量传感器,分离式电磁矢量传感器能够显著降低阵元间互耦,且非共点结构在硬件设计上更易于实现。故提出分离式电磁矢量传感器阵列结构来解决互耦问题。为实现两维高精度波达方向估计,首先提出由单个分离式全电磁矢量传感器和单个电偶极子组成的单三角阵列结构。在不增加任何阵元个数的前提下,提出一种双三角形阵列结构来实现两维波达方向的高精度估计。该矢量传感器空间结构分两个步骤设计：第一步,设计空间分离式电磁矢量传感器的空间结构使之满足矢量叉积传播矢量估计算法：第二步,在第一步基础上,设计阵列结构使之满足两维孔径扩展。上述两种阵列结构只使用六个或七个阵元,在实际的阵列雷达中往往不能满足检测概率和估计精度的需求,故提出一种稀疏均匀分离式电磁矢量传感器矩形阵列,针对该阵列提出了一种二维波达方向和极化参数的联合估计算法。但是全电磁矢量传感器中的电偶极子和磁偶极子的响应往往不一致,导致参数估计精度下降。故提出一种全电偶极子组成的三正交分离式矢量传感器阵列。上述所提阵列结构在降低阵元间互耦的同时,都采用稀疏阵列结构来扩展阵列的物理孔径,提高了波达方向估计精度。第三部分,研究双基地MIMO雷达的发射角和接收角估计问题。由于MIMO雷达的自由度等于发射阵元数和接收阵元数的Kronecker积,使MIMO雷达在提供高精度参数估计的同时,计算复杂度大大增加。此外,在双基地MIMO雷达中发射角和接收角的配对亦是一个重要问题。因此,提出一种实值ESPRIT方法和波束域求根MUSIC方法,利用全程实值操作的ESPRIT算法来降低复数域的ESPRIT算法的计算量。利用波束域的转换及求根算法来降低常规阵元域MUSIC算法的计算量。另外,从提高估计精度的角度出发,针对发射阵和接收阵均为分布式子阵的双基地MIMO雷达,研究实值双尺度ESPRIT方法来估计发射角和接收角。相对于半波长均匀分布的发射阵和接收阵组成的双基地MIMO雷达,分布式阵列能够扩展阵列的物理孔径,在不增加硬件复杂度的情况下,提高了角度估计性能。而且所提三种方法均能够实现发射角和接收角的自动配对。第四部分,研究电磁矢量传感器MIMO雷达的波达方向估计问题。鉴于MIMO雷达在参数估计方面的独特优势,考虑电磁矢量传感器在MIMO雷达中的应用,提出一种干涉式矢量传感器MIMO雷达,利用干涉发射阵列的长、短基线空间平移不变性采用双尺度ESPRIT算法获取发射角的高精度估计值；同理,利用矢量接收阵的双尺度空间平移不变特性得到高精度接收角估计值。该干涉矢量传感器MIMO阵列雷达,可同时获取MIMO雷达的波形分集和矢量传感器的极化分集,且在不增加阵元数和硬件复杂度情况下扩展有效孔径,提高了角度估计精度。另一方面,针对常规电磁矢量传感器MIMO雷达采用固定极化的发射极化方式,极化信息并没有得到充分利用,提出一种CRB最小化的发射极化优化算法来估计目标的波达方向。所提优化算法的波达方向估计精度高于采用固定极化的波达方向估计算法,并能保持固定极化波达方向估计算法的两维波达方向估计可自动配对、发射电磁矢量传感器位置可任意的优点。更多还原

【Abstract】 Polarization of electromagnetic wave is another important signal information besides amplitude, phase, frequency and waveform. The performance of radar and communication systems can be significantly improved by fully utilizing the polarization. Therefore, the polarization sensitive array consisted of electromagnetic vector sensor has broad applications in radar, communication, sonar and biomedicine. On the other hand, MIMO radar is a hot research topic in recent years with its advantages in target detection and parameter estimation. Parameter (such as DOA and polarization state angle) estimation is one of the main tasks of radar and communication systems. Therefore, this paper studies on DOA and polarization state estimation with electromagnetic vector sensor array and its applications in MIMO radar. Specific contents can be summarized as the following four parts.The first part studies the problem of DOA estimation of coherent incident signals for electromagnetic vector sensor. First, signal model is built with the scenario that completely polarized signal impinges upon traditional spatially collocating electromagnetic vector sensor, followed by reviewing of several DOA estimation algorithms based on electromagnetic vector sensor. Furthermore, no utilization of the cross-correlation information among the smoothed subarrays leads to low resolution of polarization smoothing algorithm. An improved polarization smoothing algorithm of direction-of-arrival estimation for coherent sources is proposed, which is called weighted polarization smoothing algorithm. Full use of auto-correlation and cross-correlation of the subarrays composed of six components of electromagnetic vector-sensor array is performed in weighted polarization smoothing algorithm. An equivalent covariance matrix is obtained by a weighted sum of36sub-matrixes. The derivation of theoretical formula of optimal weighting coefficients and analysis of the rank of equivalent signal covariance matrix constrained by its diagonalization are accomplished. Simulation results are presented to illustrate higher resolution and accuracy of weighted polarization smoothing against polarization smoothing.The second part proposes four structures of spatially noncollocating electromagnetic vector sensor (EMVS) array and studies its DOA and/or polarization estimation. Traditional spatially collocating EMVS array has strong mutual coupling, resulting the difficulty of the implementation of EMVS hardware, and the decline of the DOA and polarization estimation performance seriously. Spatially noncollocating EMVS (SNC-EMVS) can reduce greatly the mutual coupling and the hardware cost compared with the spatially collocating EMVS (SC-EMVS). Therefore, we propose SNC-EMVS array to solve the problem of strong mutual coupling in SC-EMVS. For two dimensional high accuracy DOA estimation, we firstly propose a new SNC-EMVS array with triangular configuration, which is composed of a single SNC-EMVS and a single dipole. Without adding sensors, we then propose a double-triangular configuration array to achieve two dimensional (2-D) high accuracy DOA estimation. The double-triangular configuration array is obtained by a two-step design. The first step aims to make the configurations of SNC-EMVS satisfy the "vector cross-product" Poynting-vector estimator. The second step focuses on extending the2-D array apertures of SNC-EMVS. Detection probability and estimation accuracy of an actual array radar often cannot be satisfied by only using six or seven sensors of the above two array structures. Therefore, we thirdly propose a sparse uniform rectangular SNC-EMVS array and a novel2-D DOA and polarization parameters estimation algorithm. But in practical applications, electric-field response and magnetic-field response in EMVS are often inconsistent, which leads to estimation performance degradation. Therefore, we fourth propose a sparse uniform rectangular SNC-EMVS array consisted of only electric-dipoles. The above four array structures not only reduce mutual coupling but also extend the2-D apertures to improve DOA estimation accuracy by using sparse structures.The third part studies the problem of DOD and DOA estimation for bistatic MIMO radar. Since the degree of freedom of MIMO radar is equal to Kronecker product between the number of transmitter and receiver, MIMO radar greatly increases computational complexity besides providing high accuracy parameter estimates. Furthermore, the pairing of DODs and DOAs in bistatic MIMO radar is also an important issue. Therefore, we propose unitary ESPRIT and beamspace root MUSIC method to reduce the computational complexity. Unitary ESPRIT uses real-valued operations throughout ESPRIT algorithm to reduce the computational complexity of the complex-valued ESPRIT algorithm. Beamspace root MUSIC uses beamspace transformation and root-like method to reduce the computational complexity of conventional MUSIC algorithm. In addition, from the perspective of improving the estimation accuracy, we propose distributed array bistatic MIMO radar and unitary two-size ESPRIT to estimate the DOD and DOA. Compared to the uniform linear array with the half-wavelength element spacing, distributed array can extend the physical aperture without increasing the hardware complexity to greatly improve the angle estimation performance. Moreover, the proposed three methods are able to achieve automatic pairing between DODs and DOAs.The fourth part studies the problem of DOA estimation for the EMVS MIMO radar. Considering the advantages of parameter estimation using MIMO radar, the EMVS is applied to MIMO radar and propose an interferometric EMVS MIMO radar. A short baseline and a long baseline of the transmitting array are utilized to obtain high accuracy DOD estimation via the two-size ESPRIT. Similarly, the high accuracy DOA estimation can be obtained by utilizing the EMVS receive array. The proposed system can obtain the waveform diversity offered by MIMO radar and the polarization diversity offered by EMVS simultaneously. Also, it is capable of extending array aperture without increasing sensors and hardware costs, which can improve the angle estimation accuracy greatly. Moreover, for the problem of the bad direction of arrival (DOA) estimation accuracy because of no utilization of the transmitted polarization information in EMVS MIMO radar, a transmitted polarization optimization algorithm based on minimizing the Cramer-Rao bound is proposed. The proposed algorithm can provide better estimation accuracy than the fixed polarization DOA estimation algorithm, and remain the advantages of the automatic pairing between the2-D DOA estimation and arbitrary placement of the transmitted electromagnetic vector sensor antennas.更多还原