| 摘要: |
| 针对在大规模多输入多输出(Multiple-Input Multiple-Output,MIMO)系统中，信道硬化现象减弱时最小均方误差(Minimum Mean Square Error,MMSE)检测算法及Richardson、Jacobi等迭代检测算法检测性能退化严重的问题，提出了一种深度检测网络，称为IPICNet,将深度学习技术和迭代并行干扰消除（Iterative Parallel Interference Cancellation,IPIC）检测算法结合。在IPICNet中，将IPIC检测算法的迭代过程展开为深度网络，并在此基础修改网络架构和添加可训练参数，同时对网络中需要使用的投影函数和损失函数进行了讨论和设计。实验结果表明，训练完成的IPICNet能有效提升IPIC检测算法的检测性能并在信道硬化现象不明显的MIMO系统中稳定工作。 |
| 关键词: 大规模MIMO 信号检测 深度学习 IPIC检测算法 |
| DOI：10.20079/j.issn.1001-893x.220405002 |
|
| 基金项目: |
|
| Deep learning network based on IPIC algorithm for massive MIMO signal detection |
| ZENG Xiangzhi,TU Yuanyuan,SHEN Bin |
| (School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China) |
| Abstract: |
| In multiple-input multiple-output(MIMO) systems,when channel hardening phenomenon weakens,the detection performance of Minimum Mean Square Error(MMSE) detection algorithm and iterative detection algorithms such as Richardson and Jacobi degrade seriously,so a deep detection network called IPICNet is proposed，which combines deep learning and Iterative Parallel Interference Cancellation（IPIC） detection algorithm.In IPICNet,the iterative process of IPIC detection algorithm is expanded into a deep network.On this basis,the network architecture is modified and trainable parameters are added.At the same time,the projection function and loss function that need to be used in the network are discussed and designed.The experimental results show that the trained IPICNet can effectively improve the detection performance of the original IPIC algorithm and work stably in MIMO systems where channel hardening is not obvious. |
| Key words: massive MIMO signal detection deep learning IPIC detection algorithm |
《电讯技术》编辑部 