基于神经网络的时变无线信道仿真  

Wireless channel simulation based on neural network

作　　者：刘留[1,2] 李慧婷 张嘉驰 李一倩 周涛 LIU Liu;LI Huiting;ZHANG Jiachi;LI Yiqian;ZHOU Tao(School of Electronic and Information Engineering,Beijing Jiaotong University,Beijing 100044,China;Beijing Key Laboratory of Mobile Computing and New Terminal,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]北京交通大学电子信息工程学院,北京100044 [2]中国科学院计算技术研究所移动计算与新型终端北京市重点实验室,北京100190

出　　处：《北京交通大学学报》2020年第2期74-82,共9页JOURNAL OF BEIJING JIAOTONG UNIVERSITY

基　　金：北京市自然科学基金(L172030)。

摘　　要：提高频谱利用率、实现超大容量传输的前提是精准掌握无线信道特性信息.研究了反馈神经网络(Back Propagation Neural Network,BPNN)在无线信道多普勒功率谱仿真中的应用.利用大量样本训练BPNN实现对时变信道进行预测.仿真结果表明,与传统方法对比,BPNN的仿真效果更接近理论值,误差更小,具备很好的容错性,同时模型输出结果的频域呈U型谱,时域自相关函数满足第一类贝塞尔函数,很好地符合Jakes模型的时频域条件.对3种方法的时间复杂度进行比较,结果表明BPNN的时间复杂度最高,牺牲了时间复杂度但换取了高精度和低误差.对3种误差逆向传播算法的仿真结果进行对比,发现列文伯格-马夸尔特(Levenberg-Marquardt,L-M)算法训练BPNN的均方误差最低,效果最佳.In order to improve the spectrum utilization rate and realize ultra-large capacity transmission,the premise is to accurately grasp the wireless channel characteristic information.This paper studies the application of Back Propagation Neural Network(BPNN)in Doppler power spectrum simulation of wireless channel.BPNN is trained by using a large number of samples to predict time-varying channels.The simulation results show that compared with traditional methods,BPNN simulation results are closer to theoretical values,with less errors and good fault tolerance.At the same time,the frequency domain of the model output results is U-shaped spectrum,and the time domain autocorrelation function satisfies the first type of Bessel function,which is in good accordance with the time-frequency domain conditions of Jakes model.The time complexity of the three methods is compared.The results show that BPNN has the highest time complexity,sacrificing the time complexity but obtaining high precision and low error.Besides,the simulation results of three error back propagation algorithms are compared.It is found that Levenberg-Marquardt algorithm has the lowest mean square error and the best effect in training BPNN.

关 键 词：无线通信 反馈神经网络 多普勒效应 时变信道 正弦波叠加法 成型滤波法 

分 类 号：U285[交通运输工程—交通信息工程及控制]