基于迁移学习的下行信道估计方法  

Model Fine-tuning Based Downlink Channel Estimation Method

作　　者：程港 CHENG Gang(South-Central Minzu University,Wuhan Hubei 430000,China)

机构地区：[1]中南民族大学,湖北武汉430000

出　　处：《信息与电脑》2023年第4期24-27,共4页Information & Computer

摘　　要：移动业务传输存在典型的非对称性,即下行数据传输需求远大于上行传输。如果基站能够利用信道信息实现调制、编码、功率自适应调整,对提升通信频谱效率、能量效率、可靠性具有重要的意义。文章提出基于迁移学习的下行信道估计算法,首先将在公开ImageNet数据集上预训练好的ResNet18网络模型参数作为模型的初始数值,其次通过模型进行微调,最后进行仿真分析。仿真结果表明,网络模型经过模型微调后,信道估计精度更高,收敛速度更快,鲁棒性更强。There is a typical asymmetry in mobile service transmission,that is,the demand for downlink data transmission is far greater than that for uplink transmission.If the base station can use the channel information to realize modulation,coding and power adaptive adjustment,it is of great significance to improve the communication spectrum efficiency,energy efficiency and reliability.This paper proposes a downlink channel estimation algorithm based on migration learning.First,the ResNet18 network model parameters pre-trained on the public ImageNet data set are used as the initial values of the model,then the model is fine-tuned,and finally the simulation analysis is carried out.The experimental simulation results show that the network model has higher channel estimation accuracy,faster convergence,and better robustness after model fine-tuning.

关 键 词：信道估计 频分双工系统 深度学习 

分 类 号：TN929.5[电子电信—通信与信息系统]