基于CBDNet的大规模MIMO系统信道估计算法  

Channel Estimation Algorithm for Massive MIMO Systems Based on CBDNet

作　　者：常文慧 许鹏 宁琦 CHANG Wen-hui;XU Peng;NING Qi(School of Electronic Information,Shenyang Aerospace University,Shenyang 110136,China;Liaoning Key Laboratory of Aerospace Information Perception and Intelligent Processing,Shenyang 110136,China)

机构地区：[1]沈阳航空航天大学电子信息工程学院,辽宁沈阳110136 [2]辽宁省空天信息感知与智能处理重点实验室,辽宁沈阳110136

出　　处：《电脑与信息技术》2025年第1期15-21,共7页Computer and Information Technology

摘　　要：针对大规模多输入多输出(Multiple-Input Multiple-Output,MIMO)系统中基于深度学习的信道估计去噪网络对不同信噪比(Signal-to-Noise Ratio,SNR)范围适应能力较弱的问题,提出了一种基于卷积盲去噪网络(Convolutional Blind Denoising Network,CBDNet)的信道估计算法。根据大规模MIMO系统在角域信道上的稀疏特性,对传统的基于卷积神经网络的CBDNet中噪声水平子网络和非盲去噪子网络的网络结构进行了重新设计,有效提升了去噪网络对不同SNR的泛化能力。此外,在CBDNet模型的训练过程中,引入了残差学习和非对称学习的损失函数,显著改善了去噪网络的鲁棒性。仿真结果表明,与现存方案相比,所提算法具有更高的归一化均方误差(Normalized Mean Squared Error,NMSE)性能。According to the weak adaptive ability to different Signal-to-Noise Ratio(SNR)ranges for deep learning based channel estimation denoising networks in massive Multiple-Input Multiple-Output(MIMO)systems,a Convolutional Blind Denoising Network(CBDNet)based channel estimation algorithm was proposed.By use of the sparse characteristics of angle-domain channel in the massive MIMO systems,the structure of the noise level subnetwork and the non-blind subnetwork in the traditional CBDNet based on convolutional neural network was redesigned,effectively enhancing the generalization ability of the de-noising network to different SNRS.In addition,during the training process in CBDNet model,the loss function of residual learning and asymmetric learning was introduced,improving the robustness of the denoising network.The simulation results show that has higher channel estimation performance than other channel estimation and denoising networks.The numerical results show that the proposed algorithm outperforms existing schemes in terms of Normalized Mean Squared Error(NMSE).

关 键 词：大规模多输入多输出 信道估计 卷积盲去噪网络 深度学习 

分 类 号：TN929.5[电子电信—通信与信息系统]