A TTSVD-Enhanced Fast and Accurate Channel Estimation Method for Dual-Polarized Massive MIMO Systems

This paper aims to perform fast and accurate downlink channel estimation for dual-polarized massive multiple-input multiple-output (MIMO) systems. Considering that both ends of the link are equipped with antenna arrays, the double directional channel matrix can be represented by a higherorder tensor. Based on the low-rankness of channel tensor due to the sparse scattering nature, we propose an ESPRIT-like channel estimation method enhanced by tensor train singular value decomposition (TTSVD). The proposed estimation method is characterized by (a) small estimation errors, which benefits from employing TTSVD to reduce noise in the channel tensor; and (b) fast computational speed, which exploits Vandermonde structure of signal subspace without multiple iterations. Simulated experiments in a 5G millimeter wave scenario show the superior estimation performance of the proposed method both on accuracy and speed. It is worth noting that the proposed method is at least 12 times faster than the state-of-The-Art.