Compressed PARAFAC-Based Parameter Estimation for mmWave Polarized Massive MIMO Systems

The electromagnetic vector sensor (EMVS) array leverages polarization diversity to estimate two-dimensional direction-of-arrival (2D-DOA) and polarization, but sacrifices computational complexity due to the multidimensional parameter estimation of high-dimensional measurement, compared to the conventional scalar sensor array (SSA). This article integrates compressive sensing (CS) theory with parallel factor (PARAFAC) analysis to develop a compressed 2D-DOA and polarization estimator for millimeter-wave (mmWave) polarized massive multiple-input multiple-output (MIMO) systems. To exploit the underlying multidimensional structure of the array measurement, a third-order tensor PARAFAC model is first formulated. Motivated by the row-orthogonal sparse (ROS) compression matrix-based tensor CS (TCS) scheme, the large PARAFAC model is compressed into a small one. The factor matrices estimation is then realized by performing PARAFAC decomposition on the small tensor. Eventually, 2D-DOA and polarization are estimated using vector cross-product and least-squares (LS) techniques. The presented scheme not only efficiently achieves satisfactory estimation accuracy but also adapts flexibly to arbitrary array geometry, making it valuable for real-world engineering scenarios. Simulation experiments validate the theoretical superiorities of the presented scheme.