基于改进DnCNN的RIS辅助毫米波系统信道估计  

Channel Estimation for RIS-assisted Millimeter Wave System Based on DnCNN

作　　者：吴颖 刘紫燕 WU Ying;LIU Ziyan(College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州大学大数据与信息工程学院,贵州贵阳550025

出　　处：《移动通信》2024年第4期86-93,共8页Mobile Communications

基　　金：贵州省联合资金资助项目“基于深度学习的行人重识别关键技术研究”(黔科合LH字[2017]7226号)。

摘　　要：RIS是第六代移动通信系统中潜在的候选技术之一。然而,由于无源的RIS缺乏信号处理能力,这给RIS辅助毫米波大规模MIMO系统的信道估计带来了挑战。为了获得更精确的信道状态信息,将信道估计转化为图像去噪问题,提出改进的DnCNN来完成信道估计任务。具体地,采用LMMSE对信道进行粗估计。融合注意力机制网络和噪声水平估计子网络对DnCNN进行改进,以提高网络对噪声的提取性能和自适应性能,实现从信道的粗估计中得到高精度信道估计值。仿真实验表明,所提算法在低信噪比下具有较好的估计性能。Reconfigurable intelligent surface (RIS) is one of the potential candidate technologies for sixth-generation wirelesscommunication system. However, due to the passive nature of RIS lacking signal processing capability, it poseschallenges to channel estimation in RIS-assisted millimeter wave (mmWave) multiple-input multiple-output (MIMO)systems. To obtain more accurate channel state information, the channel estimation is transformed into an imagedenoising problem, and an improved denoising convolutional neural network (DnCNN) is proposed to complete thechannel estimation task. Specifically, the linear minimum mean squared error (LMMSE) method is used to estimatethe channel coarsely. The DnCNN is improved by fusing the attention mechanism network and noise level estimationsub-network to enhance the network's performance on noise extraction and its adaptive performance to noise and toachieve high-precision channel estimation from the coarse one. Simulation experiments demonstrate that the proposedalgorithm exhibits a good estimation performance at low signal-to-noise ratios.

关 键 词：信道估计 RIS 去噪卷积神经网络 mmWave 注意力机制 

分 类 号：TN92[电子电信—通信与信息系统]