时变MIMO通信系统中基于CNN的改进LS信道估计  

Improved LS channel estimation based on CNN in time-varying MIMO communication systems

作　　者：安澄全[1] 高博 杨延 AN Chengquan;GAO Bo;YANG Yan(College of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《应用科技》2023年第5期66-71,共6页Applied Science and Technology

摘　　要：为了实现无线通信智能化,本文针对时变大规模多输入多输出(multiple-input multiple-output,MIMO)毫米波通信系统的信道估计进行研究。在通信系统中采用时变的Saleh-Valenzuela信道模型描述信道,该模型通过对信道的复增益引入马尔可夫过程来体现信道的时变性,可以更加精确地描述毫米波信道。针对传统的信道估计算法存在迭代次数多、收敛速度慢、计算量大等问题,本文提出一种基于卷积神经网络(convolutional neural network,CNN)的改进最小二乘(least square,LS)信道估计算法,该算法在对接收导频信号进行LS算法处理后,再将LS信道估计值输入到搭建好的卷积神经网络模型中进行处理,对时变的信道信息进行降噪和特征提取,实现对时变的Saleh-Valenzuela信道的高精度估计。该算法相较于传统的信道估计算法具有迭代次数少、速度快等优点,可以对相干时间内的时变MIMO信道进行高精度估计,实现信道估计精度的大幅度提升。In order to realize intelligent wireless communication,the channel estimation of time-varying large-scale multi-input multi-output(MIMO)millimeter-wave communication system is studied in this paper.A time-varying Saleh-Valenzuela channel model is used to describe the channel in the communication system.This model reflects the time-variability of the channel by introducing Markov process into the complex gain of the channel,which can describe the millimeter-wave channel more accurately.Aiming at the problems of traditional channel estimation algorithms,such as many iterations,slow convergence and a large amount of computation,this paper proposes an improved least square(LS)channel estimation algorithm based on convolutional neural network(CNN).After LS algorithm processing of the received pilot signals,the LS channel estimate is input into the established convolutional neural network model for processing,and noise reduction and feature extraction of time-varying channel information are carried out to achieve high precision estimation of time-varying Saleh-Valenzuela channels.Compared with the traditional channel estimation algorithms,the proposed algorithm has the advantages of less iteration times and faster speed.It can estimate the time-varying MIMO channel with high precision in the coherent time and thereby greatly improve the accuracy of channel estimation.

关 键 词：卷积神经网络 信道估计 深度学习 无线通信 时变信道 多输入多输出 压缩感知 正交频分复用 

分 类 号：TN911.5[电子电信—通信与信息系统]