人体动作与行为识别研究综述  

Research on Review on Human Action Recognition

作　　者：许芬[1] 史鹏飞 

机构地区：[1]北方工业大学电气与控制工程学院,北京100144

出　　处：《工业控制计算机》2023年第9期58-59,共2页Industrial Control Computer

摘　　要：人体动作与行为识别在智慧交通、智能安防、智能家居、人机交互、VR/AR等领域具有广泛的应用价值。由于人体动作类型繁多,且很多动作涉及与环境对象的交互,人体动作与行为识别研究存在复杂度高、易受干扰、受场景因素影响大等问题,是计算机视觉领域的一个研究难点。回顾了人体动作与行为识别研究的发展历史,对该领域的国内外研究现在进行了梳理,重点介绍了目前主流的基于图卷积神经网络的动作与行为识别研究方法。最后分析了不同方法的优缺点,并对该领域的未来发展方向进行了探讨。Human movement and action recognition has a wide range of application values in intelligent transportation,intelligent security,intelligent home,human-computer interaction,VR/AR and other fields.Since there are many types of human movements and many movements involve interaction with environmental objects,the research on human movements and action has some problems,such as high complexity,easy to be disturbed,and influenced by scene factors,which is a difficulty in the field of computer vision.This paper reviews the development history of human motion and behavior recognition research,reviews the current domestic and foreign research in this field,and focuses on introducing the current mainstream motion and action recognition research methods based on graph convolutional neural network.Finally,the advantages and disadvantages of different methods are analyzed,and the future development direction of this field is discussed.

关 键 词：动作识别 计算机视觉 深度学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]