In this paper, an algorithm of combing unitary Root-MUSIC method based on tensor mode-R and projection separation approach is proposed for multidimensional (R-D) sinusoidal parameters estimation. The model of the proposed algorithm based on tensor mode-R with the unitary matrices is firstly transferred into multiple single-sample models, and then the eigenvalue decomposition (EVD) or singular value decomposition (SVD) of a set of constructed covariance matrices is implemented to obtain the estimators of all dimensional parameters. Compared with other available methods, the computational complexity of the EVD or SVD is largely reduced due to using unitary matrices by way of transforming complex number into real-valued, moreover, the problem of parameter matching is solved by the application of the projection separation operator with tensor mode-R. Simulation results are given to demonstrate the advantage of the proposed method in terms of performance of parameter estimation as well as the computational load over several state-of-art algorithms.

中文翻译：

---

基于张量模式-R算法的酉根-MUSIC无配对参数多维正弦频率估计

本文提出了一种基于张量模式-R和投影分离方法的幺正Root-MUSIC方法相结合的多维(RD)正弦参数估计算法。提出的基于张量模式-R的幺正矩阵算法模型首先转化为多个单样本模型，然后对一组构造的协方差矩阵进行特征值分解（EVD）或奇异值分解（SVD）获得所有维度参数的估计量。与其他可用方法相比，EVD或SVD由于使用酉矩阵将复数转化为实值，大大降低了计算复杂度，并且通过投影分离算子的应用解决了参数匹配问题张量模式-R。