摘要
本文首先讨论了大规模MIMO-OFDM(Multiple-Input Multiple-Output Orthogonal Frequency Division Multiplexing)系统信道的空间相关性,提出了一种基于隐聚类假设的信道建模方法,利用概率参数模拟不同的传播环境.然后,将机器学习领域的狄利特雷过程(Dirichlet Process,DP)引入到稀疏贝叶斯学习(Sparse Bayesian Learning,SBL)模型中,建立了DP-SBL结构,在信道估计的同时挖掘并利用大规模MIMO系统所特有的隐聚类特征.接着,将DP-SBL结构应用于大规模MIMO-OFDM系统中,在因子图上利用消息传递算法推导了一种基于隐聚类和狄利特雷过程的接收机算法.最后,将本文提出的接收机算法和现有算法进行对比分析.结果表明,本文提出的接收机算法充分利用了大规模MIMO-OFDM系统特有的空间相关性,能够以较低的计算复杂度获得较强的鲁棒性和显著的性能增益.
The paper discusses the spatial correlation of channels in massive MIMO-OFDM system,and proposes a hidden clustering hypothesis to simulations different propagation environments with probability parameters.Then,the Dirichlet process(DP) in machine learning is introduced into sparse Bayesian learning(SBL) model and a DP-SBL structure is established.Consequently,the hidden clustering features of massive MIMO system are explored simultaneously in the process of channel estimation.Furthermore,the DP-SBL structure is applied to massive MIMO-OFDM systems,and a receiver algorithm based on hidden clustering and Dirichlet process is deduced by using message passing algorithm on factor graphs.Finally,we compare the proposed algorithm with the existing algorithms.Simulation results show that the proposed algorithm can exploit and utilize the spatial resources of massive MIMO-OFDM system.It can achieve remarkable performance gain with low computation complexity and strong robustness.
作者
崔建华
袁正道
王忠勇
路新华
薛琦
CUI Jian-hua;YUAN Zheng-dao;WANG Zhong-yong;LU Xin-hua;XUE Qi(School of Physics and Electronic Information,Luoyang Normal University,Luoyang,Henan 471934,China;Postdoctoral Workstation,Henan TV&Radio University,Zhengzhou,Henan,450001;School of Information Engineering,Zhengzhou University,Zhengzhou,Henan 450001,China)
出处
《电子学报》
EI
CAS
CSCD
北大核心
2019年第12期2515-2523,共9页
Acta Electronica Sinica
基金
国家自然科学基金面上项目(No.61571402)
国家青年科学基金(No.61705198)
博士后科学基金(No.2019M652576)
河南省科技攻关项目(No.182102210573)
河南省教育厅高校重点研究项目(No.19A510019)
