尺度空间特征下人机交互多姿态三维手势智能识别

打开文本图片集

关键词：三维手势识别；尺度空间特征；关节点热图；多姿态手势；人机交互；LSTM网络；体素网格；时序建模中图分类号：TN926-34；TP242 文献标识码：A 文章编号：1004-373X（2026）06-0189-05

Intelligent multi-pose three-dimensional gesture recognition for human-computer interaction under scale spatial feature

XIAOKun，GUOLingfeng²，AOSikui²，WUWei²

（1.XinjangNlUtUa;iUsitdEinoa

Abstract：Inalusion tothe diversityof gesture postures，scalevariations，and theinterferenceof complexbackgrounds in human-compuerinterationetodofitellgentult-osetree-dmesioalsturecogitionforuman-omputerteactio underscale spatialfeaturesisproposed.Thethree-dimensional gesturepointcloud isconstructedandconvertedintoabinary voxelgrid.Thegesturefeatureswithspatialdistributionfeaturesareextractedbycombiningthepyramidmulti-scalestructure andtheSIFTdescriptor.Thethree-dimensionalconvolutionalnetwork isusedtoregresiontheheatmapofthejointpointsto realizetheprecisepositioning，andthetemporal shiftmoduleandtheLSTMnetworkareintroduced tomodelthedynamic sequenceofgestures，soastorealizethereal-timeintellgentrecognitionofmulti-posegestures.Theexperimentalresultsprove thatthecomprehensiverecognition confidence levelof this method for1O types of interactivegestures can reach upto 99.68% In thethreescenarios of virtual games，office workand teaching，therecognition stabilityis97.7%，96.38%and 98.67% ，andthe anti-interference ability is 94.99 % ，93.85% and 95.98% ，which can achieve high-precision multi-pose three-dimensional gesture inteligentrecognition，andprovideeliableandnaturalgestureinteractionsupportforhuman-computerinteractionandvirtual reality applications.

Keywords：3D gesture recognition;scale spatial feature; joint point heat map;multi-posture gesture;human-computer interaction;LSTM network;voxel grid;temporal seriesmodeling

0 引言

人机交互多姿态三维手势作为新一代自然人机接口的核心，能够实现无需接触和直观、自然的指令传递与信息操控，对于提升虚拟现实沉浸感、增强现实交互效率以及复杂环境下的远程控制具有至关重要的作用。（剩余5477字）