基于注意力机制和残差网络的OFDM系统信道估计  

Channel estimation for OFDM based on attention mechanism and residual networks

作　　者：申滔 朱正发 刘受清 SHEN Tao;ZHU Zheng-fa;LIU Shou-qing(School of Electrical and Information Engineering,Changsha University of Science and Technology,Changsha 410004,China)

机构地区：[1]长沙理工大学电气与信息工程学院,湖南长沙410004

出　　处：《计算机工程与设计》2024年第12期3600-3606,共7页Computer Engineering and Design

基　　金：湖南省教育厅一般基金项目(19C0037);长沙理工大学科研创新基金项目(CLSJCX23067)。

摘　　要：为解决正交频分复用(OFDM)系统中由噪声干扰引发的导频污染问题,设计一个基于深度学习的信道估计模型CE-SERNet。将导频位置处最小二乘信道估计值当作低分辨率带噪声图像,作为网络模型输入,利用注意力机制和残差网络进行去噪和恢复高分辨率图像,实现OFDM系统的信道估计。仿真结果表明,所提网络在低导频和高导频条件下都优于现有基于深度学习的方法,相比传统的LS算法和MMSE算法,在估计精度上有较大提升,在不同的信道场景下,拥有较强的鲁棒性能。To solve the pilot pollution problem caused by noise interference in orthogonal frequency division multiplexing(OFDM)systems,a deep learning based channel estimation model called CE-SERNet was designed.The least square channel estimate at the pilot position was regarded as a low resolution image with noise,which was taken as the network input,and the attention mechanism and residual network were used to de-noise and restore the high resolution image,the channel estimation of OFDM system was realized.Simulation results show that the proposed network is superior to the existing deep learning-based methods at both low and high pilot conditions.Compared with traditional LS and MMSE algorithms,it has significant improvements in estimation accuracy and strong robustness in different channel scenarios.

关 键 词：正交频分复用 噪声干扰 导频污染 深度学习 信道估计 注意力机制 残差网络 

分 类 号：TN929.5[电子电信—通信与信息系统]