Improving Channel Estimation in IRS-Aided mmWave MIMO Systems via CNN and Sparse Bayesian Learning

Millimetre Wave (mmWave) Multiple Input Multiple Output (MIMO) systems encompassing Intelligent Reflecting Surfaces (IRS) have been recognized as one of the promising contenders to provide high spectral efficiency and improved coverage for Sixth Generation (6G) wireless networks. Channel estimation in such systems is crucial for reliable beamforming and improved system capacity. However, the task of channel estimation is challenging due to passive nature of IRS and high-dimensional channel matrix whose size proportionately increases with growing number of antennas at the base station (BS), number of IRS elements and users. It becomes even more challenging in bandwidth constrained scenarios due to limited use of pilots. In this paper, a framework combining the Deep Learning (DL) and Sparse Bayesian Learning (SBL) is developed to utilize the inherent sparsity of the cascaded channel of IRS-assisted mmWave MIMO systems in angular domain for estimating the channel in bandwidth constrained scenarios. Performance of the proposed model is compared with some of the state-of-the art algorithms. Simulation results show the superior performance of the proposed model over the counterparts in terms of Normalized Mean Square Error (NMSE).