Direction-of-arrival (DOA) estimation is essential for wireless communications, especially in the recent developed artificial intelligence (AI)-based approaches. In this paper, we investigate into the problem two-dimensional (2D) DOA using electromagnetic vector sensors (EMVS) array, and a covariance tensor-based estimator is introduced. Firstly, the array covariance matrix is formulated into a fourth-order covariance tensor. Then the higher-order singular value decomposition (HOSVD) is performed to obtain an enhanced signal subspace. Thereafter, the elevation angles are achieved via rotation invariant property. Finally, the azimuth angles are estimated by exploiting vector cross-product technique. In addition, the polarization parameters are accomplished via least squares technique, which maybe helpful to identify polarization status of the sources. The proposed covariance-based estimator can be easily extended to spatially colored noise scenario, which means it is more flexible than the state-of-the-art algorithms. Numerical simulations have been designed to show its effectiveness.

到达方向（DOA）估计对于无线通信至关重要，尤其是在最近开发的基于人工智能（AI）的方法中。在本文中，我们使用电磁矢量传感器（EMVS）阵列研究问题二维（2D）DOA，并引入了基于协方差张量的估计器。首先，将阵列协方差矩阵表述为四阶协方差张量。然后执行高阶奇异值分解（HOSVD），以获得增强的信号子空间。此后，通过旋转不变性获得仰角。最后，利用矢量叉积技术估计方位角。此外，极化参数是通过最小二乘技术完成的，这可能有助于确定光源的极化状态。所提出的基于协方差的估计器可以轻松扩展到空间彩色噪声场景，这意味着它比最新算法更灵活。设计了数值模拟以显示其有效性。