基于FAST网络的毫米波雷达端到端手势识别

打开文本图片集

中图分类号：TN957.52-34；TP18 文献标识码：A 文章编号：1004-373X（2026）01-0008-07

DOI：10.16652/j.issn.1004-373x.2026.01.002引用格式：，等.基于FAST网络的毫米波雷达端到端手势识别[J].现代电子技术，2026，49（1）：8-14.

FAST network based end-to-end gesture recognition using millimeter-wave radar

ZHENGHao，LI Haoran，PENGGuoliang，ZHENGZhipeng，HUFen，HUAN Zhan （SchoolofMicroelectronicsand ControlEngineering，Changzhou University，Changzhou 213ooo，China）

Abstract：Inviewof thecomplexities，ineficieniesandlowaccuracyofcurrentmillimeter-waveradargesturerecognition methods，thispaperproposesanFAST（Fourier-Atention-SwinTransforer）networkmodelFirstlyacomplex-valuedlinarlaer isutilizedtoconstructaFouriernetwork，andtheweightsoftheFouriernetworkareintalizedwithdiscreteFouriertransform values.Range-Dopperfeaturesareobtainedafterradar’rawdatapassing throughtheFouriernetwork．Secondly，theECA （eficientchannelatention）moduleisintroducdtocalculateframe-channelatentionweights，enhancingthecapabilitytoextract gesture features.inally，the Swin Transformerisemployedtoimprovecomputationaleficiencyandrecognitionacuracywhile expandingthereceptivefield，andalossfunctionisusedforbackpropagationanditerativeupdatesofthemodelparameters. ExperimentalresultsdemonstratethattheproposedFAST-basedend-to-endgesturerecognitionalgorithmusing millmeter-wave radarachievesanaccuracy rate of 96.46% ，showcasing advanced performance in comparison with the other mainstream algorithms.Thisstudyoffersamorestreamlinedandeficientsolutionfortheapplicationofmillimeter-waveradargesture recognition in smart homes and mobile devices.

Keywords：millimeter-waveradar;gesture recognition;human-computer interaction;deep learning；neural network;discrete Fouriertransform

0 引言

手势作为一种直观且富有表现力的人际交流方式，同样也是一种极具潜力的人机交互手段。（剩余10766字）