基于深度学习的IRS辅助无线通信系统信道估计  

Channel Estimation for IRS Assisted Wireless Communication Systems Based on Deep Learning

作　　者：张旭辉 徐岩[1] 张晓琦 ZHANG Xuhui;XU Yan;ZHANG Xiaoqi(School of Electronics and Information Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学电子与信息工程学院,甘肃兰州730070

出　　处：《无线电工程》2024年第8期1994-2001,共8页Radio Engineering

基　　金：国家自然科学基金(62063014)。

摘　　要：智能反射表面(Intelligent Reflecting Surface,IRS)能够对入射其上的信号进行一定的相位和幅度的变换,从而达到信号的精确传输,提高信号的覆盖和传输效率。但是这种优势都是在已知精确的信道状态信息(Channel State Information,CSI)的前提下才能达到。基于IRS元件的无源性,精确的CSI很难得到。针对此问题使用压缩感知(Compressive Sensing,CS)算法结合深度学习(Deep Learning,DL)的方法来解决。使用共链路结构来优化传统的压缩感知算法,能够在更低的导频开销和信噪比(Signal to Noise Ratio,SNR)下获得更好的归一化均方误差(Normalized Mean Square Error,NMSE)。将信道估计问题看作为去噪问题,把优化后的CS算法所得结果看作含有噪声的CSI,使用多重深层降噪块网络对其进一步去噪,得到更加精确的CSI。实验表明,所提算法较对比算法在相同SNR下有更好的精度。Intelligent Reflecting Surface(IRS)can perform certain phase and amplitude transformations on the incoming signals,thus achieving accurate signal transmission and improving signal coverage and transmission efficiency.However,this advantage can only be achieved when accurate Channel State Information(CSI)is known.While due to the passive nature of IRS elements,it is difficult to obtain accurate CSI.Therefore,the Compressive Sensing(CS)algorithm combined with Deep Learning(DL)is used to solve this problem.Firstly,a joint-link structure is used to optimize the traditional compressive sensing algorithms to achieve better Normalized Mean Square Error(NMSE)performance at lower pilot overhead and Signal to Noise Ratio(SNR).Secondly,the channel estimation problem is regarded as a denoising problem,the results obtained by the optimized compressive sensing algorithm are treated as CSI containing noise,and a multiple deep denoising block network is employed to further denoise the CSI to obtain more accurate CSI.Finally,the experimental results demonstrate that the proposed algorithm achieves higher accuracy than the algorithms for comparison at the same SNR.

关 键 词：智能反射表面 信道状态信息 压缩感知 共链路结构 多重深层降噪块网络 

分 类 号：TN929.5[电子电信—通信与信息系统]