基于残差学习的矿井无线信道估计的算法研究

打开文本图片集

中图分类号：TN929.4-34 文献标识码：A 文章编号：1004-373X（2025）15-0001-05

Research on algorithm for mine wireless channel estimation based on residual learning

WANGAnyi，LIMingzhu，LIXinyu，LIRuoman （CollgeofCommunicationandInformationTechnologyXi'anUniversityofienceandTechnology，Xi'an7o54Cina）

Abstract：Theexisting channel stimationneuralnetwork modelsusuallyignoreoff-diagonal elementsand temporal sequenceinformationinchannelgaincorrelationmatrices，soadeeplearningtechnologyisintroducedtoaccuratelyestmatethe channelbasedondownlinkpilotinorthogonalfrequencydivisionmultiplexing（OFDM）system.Aninnovativedeepneural networkmodelbasedonecursiveresiduallearningisproposed.Inthismodel，thesequentialdataareprocesedefectielynd thetemporalcorelationofchannelstatesiscapturedwithitsrecursivestructure.Additionall，，theintroductionofresidual connectionsefectivelymitigatesthegradientvanishingproblemcommonlyocurredindeep learning，whichenhancesthe trainingeffectoftheodelsignificantly.Furthermore，tisstudyintegatestheSE（queze-and-excitation）atentionmhaism， whichenablesthenetwork toadaptivelyadjustitsfocusondiferentchannelmatrices，therebyimprovingfeatureextractionand classification eficiency.Evaluationsof therecursiveresidual network-basedchannelestimationmodelunderthe3GPPchanel modeldemonstratethatthemethodoutperformsthetraditionalleast-squaresmethodsandtheReEsNetchannelestimation algorithm in terms of channel estimation error.

Keywords：mine communication;deep learning;residual learning;atention mechanism; OFDMsystem;time seriesdata; gradient vanishing problem;3GPP channel model

0 引言

在煤矿智能化建设[1-3中，稳定可靠的宽带无线通信系统发挥了至关重要的作用。（剩余6305字）