Super-Resolution Block-Sparse Channel Estimation Over Uplink M-MIMO 5G Mobile Wireless Networks

In this paper, we develop a novel super-resolution block-sparse channel estimation approach for uplink massive multi-input multi-output (MIMO) based 5G mobile wireless networks. We first introduce a pattern-coupled Bernoulli-Gaussian (PC-BG) prior to characterize block-sparse channels resulting from a small number of scatterers with a small range of angular spread in the angular domain. Then, we propose a Bayesian inference method to infer the channel vector as well as its hyperparameters associated with the PC-BG prior. Specifically, the proposed algorithm is developed within an expectation-maximization (EM) framework and integrated with the generalized approximate message passing (GAMP) technique of approximating the intractable posterior distribution. Finally, instead of adopting some predefined basis/dictionary, such as a discrete Fourier transform (DFT) basis, we propose to learn the offgrid gap between sampled grid-point and true angle of arrivals (AoAs) iteratively for better sparse channel representation, which thus can improve the sparse channel recovery performance. Our numerical analyses validate and evaluate the effectiveness of our proposed scheme.