An Improved PARAFAC Estimator for 2D-DOA Estimation Using EMVS Array

The topic of direction-of-arrival (DOA) estimation has attracted extensive attention in wireless communications, radars, sonars, etc. Compared to the traditional scalar sensor, electromagnetic vector sensor (EMVS) is attractive since it provides two-dimensional (2D) DOA estimation and additional polarization information of the incoming signals. However, existing algorithms are only suitable for Gaussian white noise scenario. In this paper, we investigate into DOA estimation using EMVS array with spatially colored noise, and a parallel factor (PARAFAC) estimator is proposed. Unlike the conventional direct PARAFAC algorithm, the covariance tensor-based PARAFAC model is considered. For fast PARAFAC decomposition purpose, the fourth-order covariance PARAFAC tensor is rearranged into a third-order PARAFAC tensor, so that the existing COMFAC algorithm is available and thus the factor matrices are estimated. Thereafter, the elevation angle estimation is accomplished via least squares (LS) technique, and the azimuth angles are estimated via vector cross-product. Besides, the polarization status of the source signal can be estimated via LS approach, which may be helpful to identify weak signals. Our estimator is flexible since it can be easily extended to nonuniform array and spatially colored noise scenario. Moreover, it offers better estimation performance than the traditional PARAFAC algorithm in the presence of colored noise. Detailed analyses concerning identifiability, complexity as well as Cramer–Rao bound (CRB) are provided. To show the effectiveness of the proposed estimator, numerical simulations have been designed.