首页 | 本学科首页   官方微博 | 高级检索  
     

基扩展模型下基于LSTM神经网络的时变信道预测方法
引用本文:聂倩,杨丽花,呼博,任露露. 基扩展模型下基于LSTM神经网络的时变信道预测方法[J]. 系统工程与电子技术, 2022, 44(9): 2971-2977. DOI: 10.12305/j.issn.1001-506X.2022.09.33
作者姓名:聂倩  杨丽花  呼博  任露露
作者单位:南京邮电大学江苏省无线通信重点实验室, 江苏 南京 210003
基金项目:江苏省科技厅自然科学基金(BK20191378);江苏省高等学校自然科学研究面上项目(18KJB510034);批中国博士后科学基金特别资助项目(2018T110530);国家重大研究计划重点项目(92067201)
摘    要:针对高速移动正交频分复用系统, 提出了一种基扩展模型(basis expansion model, BEM)下基于长短期记忆(long short-term memory, LSTM)神经网络的时变信道预测方法。为了降低传统BEM的建模误差, 根据高速移动环境中不同列车在相同位置处的无线信道具有强相关性的特点, 首先基于历史时刻的信道状态信息获取最优的基函数, 并利用该基函数对信道进行建模。然后, 通过LSTM神经网络对信道基系数进行线下训练与线上预测来获取未来时刻信道信息, 大大降低了计算复杂度。在线下训练中, 将网络的逼近目标设置为信道估计值, 而不是理想的信道信息, 以增强预测模型的实用性。仿真结果表明, 相比现有方法, 新方法的计算复杂度较低, 且预测精度较高。

关 键 词:高速移动  长短期记忆神经网络  基扩展模型  时变信道预测  
收稿时间:2021-10-12

Time-varying channel prediction method based on LSTM neural networks under basis expansion model
Qian NIE,Lihua YANG,Bo HU,Lulu REN. Time-varying channel prediction method based on LSTM neural networks under basis expansion model[J]. System Engineering and Electronics, 2022, 44(9): 2971-2977. DOI: 10.12305/j.issn.1001-506X.2022.09.33
Authors:Qian NIE  Lihua YANG  Bo HU  Lulu REN
Affiliation:Jiangsu Key Laboratory of Wireless Communication, Nanjing University of Posts and Telecommunications, Nanjing 210003, China
Abstract:For high-speed mobile orthogonal frequency division multiplexing system, a time-varying channel prediction method based on long short-term memory (LSTM) neural network under basis expansion model (BEM) is proposed. To reduce the modeling error of the traditional BEM, according to the strong correlation characteristics of the wireless channel at the same location for the different trains in a high-speed mobile environment, the optimal basis function is obtained by the channel sate information of historical time, and it is used to model the channel. Then, the channel information at the future time is obtained by the offline training and online prediction of the channel base coefficient via LSTM neural network, which greatly reduces the computational complexity. In offline training, to enhance the practicality of the prediction model, the channel estimation, rather than the ideal channel information, is set to the approximation objective of the network. The simulation results show that the proposed method has lower computational complexity and better prediction accuracy comparing with the existing methods.
Keywords:high-speed mobile  long short-term memory (LSTM) neural network  basis expansion model (BEM)  time-varying channel prediction  
点击此处可从《系统工程与电子技术》浏览原始摘要信息
点击此处可从《系统工程与电子技术》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号