Low Complexity Transform Coding for Millimeter Wave MIMO CSI Compression

摘要

Channel state information (CSI) obtained at the user equipment (UE) can be made available at the base stations (BS) using limited feedback strategies. The design of feedback techniques is critical in millimeter wave (mmWave) multiple-input multiple-output (MIMO) systems, due to the high dimension of the channel matrix. In this paper, we assume that the mmWave MIMO channel has been estimated, and develop a compression method that allows trading off rate and distortion. The proposed method, which exhibits low complexity, is based on transform coding and uses a customized significance map to encode the relevant coefficients without using a fixed ordering scheme, such as the zig-zag scanning used in image coding. We show the performance gain of our method compared to previous algorithms using several datasets of MIMO channels obtained with ray tracing simulations, for mobile and fixed access scenarios.