Channel Estimation for Intelligent Reflecting Surface Enabled Terahertz MIMO Systems: A Deep Learning Perspective

Terahertz (THz) communications have been recognized as a promising technology to provide sufficient spectrum resources and ultra-high data rate for sixth generation (6G) wireless communication networks. To compensate the coverage hole caused by the propagation features of THz waves, an intelligent reflecting surface (IRS) is proposed to create the controllable propagation environment. However, the channel acquisition is particularly complicated for the IRS enabled THz multiple-input multiple-output (MIMO) system, since the IRS is lack of the signal processing capability. To this end, we firstly convert the channel estimation problem into a spare recovery problem by utilizing the sparsity nature of the THz channel. Then a deep learning based channel estimation (DL-CE) scheme is developed to solve the sparse recovery problem by revealing the relationship between the received signals and path gains. Simulation results demonstrate that, in contrast with classic compressed sensing based methods, our proposed DL-CE method achieves a better recovery performance and greatly decreases the computational complexity.