中文摘要：

针对极化敏感阵列信号波达方向（direction of arrival, DOA）估计问题，提出了一种基于塔克张量域序贯截断高阶奇异值分解的正则极化旋转不变参数估计（Tucker tensor based regularized polarimetric estimationof signal parameters via rotational invariance technique, trpESPRIT）方法。首先对阵列接收信号进行塔克张量建模，之后通过序贯截断高阶奇异值分解获得塔克张量域信号子空间，最后利用多旋转不变子空间幅相关系获得信号DOA估计。相比于传统矩阵建模方法，塔克张量建模更便于组织多维数据结构，实现高维的数据匹配操作，而序贯截断高阶奇异值分解则可以获得更高的信号子空间估计精度以及后续的DOA估计。仿真结果表明，trpESPRIT方法较之常规矩阵方法和矢量方法可以更好地抑制噪声，具有更高的信号DOA估计精度，在低信噪比和低快拍条件下仍然具有良好的分辨能力。

英文摘要：

The problem of direction of arrival （DOA） estimation using a polarization sensitive array is ad- dressed. A Tucker tensor based regularized polarimetric estimation of signal parameters via rotational invariance technique （trpESPRIT） is proposed by using sequentially truncated higher-order singular value decomposition （STHOSVD）. In the method, the Tucker tensor model for the polarimetric measurements is firstly established, then the signal subspace is obtained by using STHOSVD, and the DOAs of signals are finally obtained by using multiple rotationally invariant subspace amplitude and phase information. Compared with the traditional matrix methods, the Tucker tensor modeling scheme is more convenient for the characterization of the multidimensional data structure and the multidimensional data matching operation. The STHOSVD can be used to obtain more accurate signal subspace and the subsequent DOA estimation. The simulation results show that, compared with the matrix and vector methods, the trpESPRIT has a higher noise suppression capability and a higher DOA esti- mation precision. Under the condition of low signal-to-noise ratio and few snapshots, the trpESPRIT is still ob- served to have a good resolution.