Semi-Blind Channel Estimation for Massive MIMO Systems Exploiting Low-Rank Matrix Reconstruction

In order to improve channel estimation for massive multiple-input multiple-output (MIMO) systems, we propose a novel method to mitigate the effect of additive Gaussian noise on the received signal utilizing the fundamental assumption for massive MIMO systems that the number of antennas at the base station (BS) is much larger than the number of several users. We use the asymptotic singular value distribution of a low-rank matrix plus Gaussian noise matrix to reconstruct the signal via the singular value decomposition (SVD) of the received signal. We refer to this new method as low-rank SVD (LRSVD). Simulation results confirm that applying LRSVD can improve the channel estimation and data detection performance significantly for low-to-moderate signal-to-noise ratios, in particular for the semi-blind methods using independent component analysis (ICA) which are sensitive to noise and require significantly noise suppression.