Covariance Estimation Of Millimeter Wave Channels Using Sparse Signal Recovery Algorithms In A Hybrid MIMO Architecture

In this paper, the channel covariance estimation of a single mobile station (MS) in a narrowband millimeter wave (mmWave) communication system was addressed. The communication system worked in time division duplex (TDD) mode and the channel covariance was estimated in the uplink communication. The base station (BS) had multiple antennas with a hybrid architecture of radio frequency (RF) chains made up of analog and digital combiners, while the MS had a single antenna. The investigated system model assumed the shared combining matrix scheme where the same combining matrix was used across multiple coherence blocks of the mmWave channel. The application of the sparse signal recovery algorithms including the simultaneous orthogonal matching pursuit (SOMP), the multiple response sparse Bayesian learning (MSBL), and the correlated sparse Bayesian learning (CSBL) to the system model were shown. The algorithms were evaluated numerically, and their normalized mean square error (NMSE) performances were compared against the benchmark oracle minimum mean square error (MMSE) estimator in multiple scenarios of varying number of RF chains at the BS and sparsity ratios for modeling the mmWave channel. The numerical results indicated that the CSBL algorithm provided the NMSE results closest to that of the oracle MMSE estimator in all the scenarios.