SBL-Based Hybrid Precoder/ Combiner Design for Power and Spectrally Efficient Millimeter Wave MIMO Systems

Abstract

This work proposes a novel sparse Bayesian learning (SBL)-based hybrid precoder/ combiner design scheme for millimeter wave (mmWave) MIMO systems. Towards this end, a multiple measurement vector (MMV) based sparse signal recovery problem is developed that maximizes the mutual information by approximating the hybrid precoder to the ideal digital precoder. A unique aspect of the proposed SBL-based scheme is that the resulting hyperparameter estimates can be used to activate the minimum number of RF chains required to approximate the ideal digital precoder/ combiner, thus enabling one to leverage the time-varying multipath profile of the underlying mmWave MIMO channel. This feature coupled with the improved ability of SBL for sparse signal recovery leads to a significantly enhanced power and spectral efficiency of the proposed scheme in comparison to the conventional schemes that activate a fixed number of RF chains and data streams, irrespective of the multipath profile of the mmWave MIMO channel. Simulation results demonstrate the improved efficiency of the proposed scheme in comparison to the existing schemes and also the resulting reduction in the average number of RF chains employed.