| 摘要: |
| 对DetNet结构进行改进以提升大规模MIMO（Multiple-Input Multiple-Output）信号检测性能。首先，去除输入端的冗余向量简化网络中每个检测单元的结构;其次，为了进一步提升网络性能，借鉴随机森林在每个决策树的输入端引入随机性的思想，通过复制网络将原网络中的检测单元扩充为两个，构造成孪生网络的结构，并在其输入端设置不同的初值向量。仿真结果表明，结构优化后的网络比原DetNet具有一定的性能提升。 |
| 关键词: 大规模MIMO 信号检测 深度学习 孪生网络 迫零算法 |
| DOI：10.20079/j.issn.1001-893x.210408001 |
|
| 基金项目:国家自然科学基金资助项目(61501315，62072325) |
|
| Modified DetNet massive MIMO detector |
| LI Suyue,JIA Peng,WANG Anhong |
| (School of Electronic Information Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,China) |
| Abstract: |
| The DetNet structure is modified to enhance the massive multiple-input multiple-output(MIMO) signal detection performance.Firstly,the structure of each detection unit in the network is simplified by removing the redundant vectors at the input.Secondly,to further improve the network performance,the idea of random forest is used which can introduce randomness at the input of each decision tree.Specifically,the detection unit in the original network is doubled by replication to construct the siamese network and different initial value vectors are introduced at the input.The simulation results show that the optimized network achieves a certain performance improvement over the original DetNet. |
| Key words: massive MIMO signal detection deep learning siamese network zero forcing algorithm |
