Intelligent Near-Field Channel Estimation for Terahertz Ultra-Massive MIMO Systems
Anho Lee, Hyungyu Ju, Seungnyun Kim, Byonghyo Shim
Abstract
The terahertz (THz) communication systems as-sisted by ultra-massive (UM) number of antennas have been considered as a promising solution for future 6G wireless communications. As a means to overcome the severe propagation loss arising from THz band and thus achieve high beamforming gain, ultra-massive multiple-input-multiple-output (UM-MIMO) sys-tems have received much attention. To realize highly directional communications, acquisition of accurate channel state information is essential but the channel estimation techniques designed for the ideal far-field channel result in severe performance loss in the near-field region. In this paper, we propose an intelligent near-field channel estimation technique for THz UM-MIMO systems. To be specific, we extract the channel parameters, i.e., angles, distances, time delay, and complex gains, by exploiting the convolution neural network (CNN), a deep learning network specialized in capturing the spatially correlated features from the input data. From the simulation results, we demonstrate that the proposed scheme outperforms the conventional channel estimation schemes in terms of the bit error rate (BER) and the pilot overhead reduction.