Joint Matrix Completion and Randomized SVD Channel Estimation of the mmWave Massive MIMO

Abstract

Beamforming technology is usually applied to millimeter wave(mmWave) massive MIMO systems to improve transmission gain, spectrum efficiency, and anti-interference ability. The effective acquisition of channel state information(CSI) is necessary for beamforming technology. However, the low signal-to-noise(SNR) environment of beamforming and the high variability of the mmWave channel make it challenging to acquire CSI. Recently, some researches on the sparsity of the mmWave channel in the angular domain have been carried out. And it's proved that the mmWave channel exhibits a joint sparse and low-rank structure where the rank is far smaller than the sparsity level of the channel. But numbers of training symbols are still needed in these works to achieve satisfactory performance. In this work, we propose a novel channel estimation algorithm based on matrix completion which jointly utilizes the sparsity and low-rank characteristics of the mmWave channel, and the randomized singular value decomposition(SVD) is introduced to accelerate the process of the algorithm. Simulation results demonstrate that the proposed channel estimation algorithm can show more accurate performance under the condition of shorter training sequences and a low SNR environment in the mmWave massive MIMO systems.