基于深度学习的独居老人行为识别系统设计  

Design of a Deep Learning-Based Behavior RecognitionSystem for Elderly Living Alone

作　　者：刘威 杜芃森 陈兴文 Liu Wei;Du Pengsen;Chen Xingwen(School of Information and Communication Engineering,Dalian Minzu University,Dalian Liaoning 116600,China)

机构地区：[1]大连民族大学信息与通信工程学院,辽宁大连116000

出　　处：《山西电子技术》2025年第2期59-61,共3页Shanxi Electronic Technology

摘　　要：针对当今人口老龄化和独居老人的现象,设计了基于深度学习的独居老人行为识别系统,旨在通过优化YOLOv5s目标检测算法并结合轻量化OpenPose模型算法,实现对老年人日常行为的准确、实时监测,能够及时发现异常行为并采取相应的预警措施。实验证明,该系统对行为检测准确率平均值达到95%,检测速度达到27fps,从而验证了系统的有效性和可行性。In response to the phenomenon of aging population and elderly living alone,a deep learning-based behavior recognition system for elderly living alone is designed.The system aims to achieve accurate and real-time monitoring of daily behaviors of elderly people through optimizing the YOLOv5s object detection algorithm and combining with the lightweight OpenPose model algorithm so as to detect abnormal behaviors in a timely manner and take corresponding early warning measures.Experiments have shown that the system achieves an average accuracy rate of 95%for behavior detection and a detection speed of 27fps,thus verifying the effectiveness and feasibility of the system.

关 键 词：YOLOv5s 人体姿态估计 行为检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]