电讯技术2026，Vol.66Issue(5)：747-754,8.DOI:10.20079/j.issn.1001-893x.250110002
RIS辅助通信系统中的通道注意力残差网络信道估计
Channel Attention Residual Network for Channel Estimation in RIS-assisted Communication Systems
摘要
Abstract
A channel estimation method based on channel attention residual network(CA-ResNet)is proposed to address the challenges caused by path loss in communication systems and the high pilot overhead in traditional reconfigurable intelligent surface(RIS)channel estimation methods.The method combines least squares(LS)estimation with deep learning techniques to reconstruct low-resolution channel matrices into high-resolution ones.To reduce pilot overhead,a strategy of grouping RIS reflection elements is introduced,where each group of elements shares the same reflection coefficient.CA-ResNet extracts key features through residual modules and optimizes channel weights using a dual-pooling channel attention(DPCA)module.Simulation results show that with a grouping number of 4,the proposed method reduces normalized mean squared error by 2.28～2.83 dB and 1.21～1.33 dB compared with enhanced deep super-resolution(EDSR)and global attention residual network(GARN),respectively.关键词
信道估计/智能反射面/深度学习/通道注意力机制Key words
channel estimation/reconfigurable intelligent surface/deep learning/channel attention mechanism分类
信息技术与安全科学引用本文复制引用
杨黎明,高晓敏..RIS辅助通信系统中的通道注意力残差网络信道估计[J].电讯技术,2026,66(5):747-754,8.基金项目
重庆市自然科学基金创新发展联合基金(中国星网)(CSTB2023NSCQ-LZX0114) （中国星网）
重庆市自然科学基金面上项目(cstc2021jcyj-msxmX0454) （cstc2021jcyj-msxmX0454）
