DOI:10.20079/j.issn.1001-893x.210408001引用格式:李素月,贾鹏,王安红.改进的DetNet大规模MIMO检测器[J].电讯技术,2023,63(2):220-225.[LISY,JIAP,WANGAH.ModifiedDetNetmassiveMIMOdetector[J].TelecommunicationEngineering,2023,63(2):220-225.]改进的DetNet大规模MIMO检测器*李素月,贾鹏,王安红(太原科技大学电子信息工程学院,太原030024)摘要:对DetNet结构进行改进以提升大规模MIMO(Multiple-InputMultiple-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;迫零算法开放科学(资源服务)标识码(OSID):微信扫描二维码听独家语音释文与作者在线交流享本刊专属服务中图分类号:TN911.23文献标志码:A文章编号:1001-893X(2023)02-0220-06ModifiedDetNetMassiveMIMODetectorLISuyue,JIAPeng,WANGAnhong(SchoolofElectronicInformationEngineering,TaiyuanUniversityofScienceandTechnology,Taiyuan030024,China)Abstract:TheDetNetstructureismodifiedtoenhancethemassivemultiple-inputmultiple-output(MIMO)signaldetectionperformance.Firstly,thestructureofeachdetectionunitinthenetworkissimplifiedbyremovingtheredundantvectorsattheinput.Secondly,tofurtherimprovethenetworkperformance,theideaofrandomforestisusedwhichcanintroducerandomnessattheinputofeachdecisiontree.Specifically,thedetectionunitintheoriginalnetworkisdoubledbyreplicationtoconstructthesiamesenetworkanddifferentinitialvaluevectorsareintroducedattheinput.ThesimulationresultsshowthattheoptimizednetworkachievesacertainperformanceimprovementovertheoriginalDetNet.Keywords:massiveMIMO;signaldetection;deeplearning;siamesenetwork;zeroforcingalgorithm0引言大规模MIMO(Multiple-InputMultiple-Output)技术是5G通信系统的关键技术之一。在大规模MIMO通信系统中,基站(BaseStation,BS)部署大量的天线为更多用户同时提供服务。大规模MIMO可以实现较高的分...