智能系统学报2025，Vol.20Issue(4)：763-775,13.DOI:10.11992/tis.202407012
基于惯性测量单元的人体运动意图识别方法:现状与挑战
Human motion intention recognition method based on inertial measurement unit:current situation,and challenges
摘要
Abstract
Human activity recognition(HAR)utilizes wearable computing,machine learning,and other technologies to identify and understand human behaviors,which remarkably enhances current human living standards in areas such as behavior tracking,health monitoring,and human-computer interaction.Inertial sensors have increasingly become the mainstream devices in wearable computing due to their highly compact size,low cost,and stable signal characteristics.Consequently,much research in the HAR field employs inertial signals as data sources and applies deep learning al-gorithms to address challenges in data utilization,privacy protection,and model deployment.This paper systematically introduces deep learning approaches for HAR and categorizes and summarizes existing work,and comprehensively ana-lyzes current advancements,development trends,and key challenges.First,this paper introduces mainstream wearable devices used in HAR research and their data modalities,and details the characteristics of each modality.Second,this pa-per compiles commonly used HAR datasets in recent years and summarizes the data modalities,sensor placements,movement types,and citation frequencies within each dataset.Furthermore,the paper reviews the progress of several deep learning methods commonly applied in the HAR field from the perspectives of algorithm characteristics and ap-plication scenarios.Finally,this paper discusses the challenges currently confronting deep learning in the HAR field and the potential solutions.关键词
人体行为识别/深度学习/惯性传感器/普适计算/数据隐私/模型部署/迁移学习/数据质量Key words
human activity recognition/deep learning/inertial sensors/ubiquitous computing/data privacy/model de-ployment/transfer learning/data quality分类
信息技术与安全科学引用本文复制引用
衣淳植,贾翊丞,姜峰,王修来..基于惯性测量单元的人体运动意图识别方法:现状与挑战[J].智能系统学报,2025,20(4):763-775,13.基金项目
江苏省科技计划项目(BE2021086) （BE2021086）
中央引导地方科技发展专项(2024ZYD0266). （2024ZYD0266）
