重庆邮电大学学报（自然科学版）2025，Vol.37Issue(2)：232-240,9.DOI:10.3979/j.issn.1673-825X.202401170014
非完美反馈链路下可变压缩率CSI反馈方法
Changeable compression ratio CSI feedback method under imperfect feedback links
摘要
Abstract
In frequency division duplex(FDD)massive multiple input multiple output(MIMO)systems,a channel state information(CSI)feedback method based on changeable compression rate and denoising neural network is proposed to ad-dress the fixed compression rate and asymmetric FDD feedback link of current deep learning-based CSI feedback methods.This method adopts an autoencoder structure,which compresses CSI using a fixed compression rate encoder at the user end.Then,feedback constraints are used to limit the length of the feedback codeword.When the base station receives the feed-back codeword,it first fills the codeword with zero through feedback constraints,then uses a denoising unit to denoise it,and finally uses a decoder to reconstruct the denoised codeword into a CSI matrix.Simulation in a clustered delay line(CDL)channel environment shows that the proposed method can reduce storage overhead by about 48%compared to Csi-Net,and effectively suppress the impact of feedback errors on channel state information reconstruction.关键词
大规模多输入多输出/信道状态信息/压缩反馈/深度学习Key words
massive MIMO/channel state information/compress and feedback/deep learning分类
电子信息工程引用本文复制引用
黄凤翔,段红光,毛翔宇..非完美反馈链路下可变压缩率CSI反馈方法[J].重庆邮电大学学报（自然科学版）,2025,37(2):232-240,9.基金项目
重庆市自然科学基金项目(CSTB2022NSCQ-MSX1125) Natural Science Foundation of Chongqing(CSTB2022NSCQ-MSX1125) （CSTB2022NSCQ-MSX1125）
