传感技术学报2025，Vol.38Issue(9)：1597-1605,9.DOI:10.3969/j.issn.1004-1699.2025.09.008
一种基于多源信息融合的跌倒检测研究
Research on Fall Detection Based on Multi-Source Information Fusion
摘要
Abstract
Single information source is usually used in the traditional fall detection method,which leads to problems such as low detec-tion accuracy,weak algorithm robustness,and poor stability in the process of practical application.A fall detection method based on in-ertial sensor data and sound data is proposed.The method is targeted at wearable devices,it adopts high-performance lightweight neural networks(Ghost-InertialNet,Ghost-SoundNet)and information fusion for fall detection.In this algorithm,the primary detection method is based on inertial sensor(accelerometer,gyroscope)data,while the method based on fall sound data is used as an auxiliary means.The experimental results show that the algorithm achieves satisfactory detection accuracy and has a small number of model parameters,which meets the requirements for deployment on resource-constrained wearable devices.Meanwhile,incorporating the sound model gives the overall algorithm higher accuracy and reliability compared to the single information source approach.关键词
跌倒检测/多源信息融合/惯性传感器数据/声音数据/加权平均法Key words
fall detection/multi-source fusion/inertial sensor data/sound data/weighted average method分类
信息技术与安全科学引用本文复制引用
方堃,潘巨龙,项睿涵,李玲艺..一种基于多源信息融合的跌倒检测研究[J].传感技术学报,2025,38(9):1597-1605,9.基金项目
浙江省基础公益研究计划项目(LGF21F020017) （LGF21F020017）
