本文针对宽频段窄带相干信号源，充分利用多重信号分类和旋转不变两大技术，结合一定的时域信息，提出了一种高精度的信号频率和波达方向同时估计方法。该方法通过构造平滑波达方向矩阵进行参数估计，无需搜索过程和配对处理。计算机仿真验证了本文所提出方法的有效性。

目的 研究空间窄带信号源的频率和波达方向（DOA）的同时估计方法。方法 针对宽频段窄带信号，根据阵列接收信号，利用时间平移构造伪数据矩阵，同时，将伪数据矩阵信号子空间进行旋转，结合旋转不变技术，通过对构造的所谓信号空间矩阵进行特征分解，提取导向矢量矩阵和旋转因子中所包含的角度和频率信息，从而直接获得二维参数估计。结果 计算机仿真表明该法在中等信噪比下是有效的，且无需搜索过程和配对处理。

本文对基于一维ESPRIT方法联合估计空间源波达方向和频率的传统方法进行了深入分析，指出了传统旋转因子匹配对准方法存在的问题，给出了一种改进的匹配对准方法。仿真表明，这一改进方法行之有效，并可推广到其它用一维方法解决二维估计问题的应用场合。

基于四阶累积量的几何解释，本文提出了一种新的旋转不变（ESPRIT）波达方向估计方法，它通过对所构造出的虚拟旋转不变矩阵对进行广义特征分解直接得到角度估计，且与基于二阶统计量的旋转不变方法比较阵列有效孔径扩展了一倍。计算机仿真验证了本文所提方法的有效性。

A new method is proposed to provide the simultaneous estimation of both the Doppler frequency and the Direction-of-Arrival (DOA) of multiple signals received by a pulse Doppler phased array radar. Based on the concept of multiple invariance, this approach automatically pairs the Doppler frequencies and DOAs by computing them directly from the eigenvalue decomposition of a so-called signal space-time matrix (SSTM). Simulation results show that the proposed method is effective.

研究如何简化三极子天线导向矢量平凡（秩-1）模糊问题的分析。基于对两个矢量进行相同的满秩变换并不改变两者之间的线性无关性这一事实，证明了坐标系旋转并不影响导向矢量平凡模糊问题的分析。通过坐标系的旋转使分析信号入射方向与新坐标系的x 轴重合，而潜在模糊信号传播矢量落于相应xy平面内，使得三极子天线信号导向矢量亦即极化-角度域相干结构具有简单形式，简化了相应平凡 模糊问题的研究。理论分析和计算机仿真均表明，利用简便方法所得到的关于平凡模糊发生条件的结论与文献[1]中基于最小均方误差算法的自适应三极子天线信干比恶化的几种情况完全吻合。

In this paper, we propose some preprocessing techniques that improve the performance of the LSF based method for DOA estimation in the presence of rank deficiency in time correlation matrix frequently occurred in practical application. In these methods, the degrees of freedom are reduced so as to fulfill the ill-conditioned optimization problem with several schemes such as temporal smoothing, subspace constraint, DFT beamspace transformation and diagonal loading being taken. Simulation results verify the efficacy of the proposed modifications.

This contribution presents two versions of so-called regularized ESPRIT to jointly estimate 2-D directions-of-arrival (DOAs) and polarizations of multiple non-Gaussian independent polarized sources by using one electromagnetic (EM) vector sensor in the sense of regularized least squares (RLS) and regularized total least squares (RTLS), respectively. In the given two methods, second-order and fourth-order statistics are combined to restore the six-dimensional polarization-angular coherence structure for one EM vector sensor from which unambiguous joint estimation of 2-D DOAs and polarizations can be further obtained via a vector cross-product or a multi-phase-comparing procedure. Simulation results are also included to validate the proposed methods.

The output signal to interference-plus-noise ratio (SINR) of the minimum variance (MV) beamformer with one electromagnetic (EM) vector sensor in the presence of non-identical electric and magnetic noise power, which was not studied in Nehorai's earlier work [1], is derived in this aper. The state of polarization of the desired signal and interference under consideration are both assumed to be completely polarized. It is shown that the output SINR converges to that of a tripole antenna and a fairly high level near to infinite as the ratio of magnetic noise power and electric noise power approaches infinity and zero, respectively. It also has different sensitivity to the mismatch of sensor noise power for different types of states of polarization of the desired signal and the interference, e.g., in the circular, elliptical, and linear manner, and different DOA separation between the desired signal and the interference.

It is shown in the paper how a linear array of diversely polarized antennas with one pair of identical sensors is used to obtain closed-form unambiguous estimation of 2-D DOA and polarization. Spatial phase information together with weighted 3-D polarization-angular coherence structure (PACS) are first recovered with fourth-order cumulants manipulation via a new 2-D ESPRIT variant. Spatial filtering is performed to obtain the scaled PACS, from which closed-form 2-D DOA and polarization estimates can be derived with only quadrant ambiguity involved. The undesired quadrant ambiguity can be further resolved by using the acquired estimate of spatial phase factor.