Super-Resolution Wide-Beam Training for Multiuser mmWave Massive MIMO Systems

In this paper, we investigate beam training for multiuser millimeter wave massive MIMO. To reduce the training overhead, a super-resolution wide-beam training scheme including three stages is proposed. In the first stage, we perform beam sweeping based on a wide-beam codebook, where a multipath detection method based on extreme point detection is proposed to generate candidate wide-beam pairs for multiuser beam allocation. In the second stage, we propose a narrow-beam prediction method to refine the allocated wide-beam pair. In the third stage, a super-resolution angle estimation method which can break through the resolution limitation is proposed to further calibrate the channel angle-of-arrival and angle-of-departure of the predicted narrow-beam pair. Simulation results demonstrate that the proposed scheme can approach the performance of the existing beam sweeping with only a quarter of the training overhead.