MAGLN: Multi-Attention Graph Learning Network for Channel Estimation in Multi-User SIMO

Channel estimation is one of the fundamental topics in practical multi-antenna systems. With the progress of artificial intelligence, deep learning (DL)-based schemes have presented enormous the potential for performance and efficiency. In this paper, we propose an attention-aided approach to achieve channel estimation for multi-user single input multiple output (SIMO) system. Specifically, the multi-attention graph learning network (MAGLN) is conducted to estimate the uplink channel, which concentrates on the partial more important information in the different dimensions. The channel attention and graph attention mechanisms are adopted to enhance the quality of extracted features and finally output the estimated channel information. Numerical results show that the proposed scheme has better estimation performance compared with the traditional algorithms and other candidate DL-based architectures.