摘要
针对基于可穿戴设备传感器、视频分析和环境传感器的跌倒检测方法储存资源受限、计算资源消耗大和精度低的缺点,提出了一种基于姿态估计的静态图像跌倒检测方法。利用卷积神经网络提取人体的姿态估计,通过人体的姿态估计判断出人体是否为跌倒状态,利用分类网络进行跌倒姿态的验证。实验结果表明,基于姿态估计的静态图像跌倒检测方法识别率高、计算资源消耗低、速度快。
Aiming at the problem that fall detection method of wearable device sensor,video analysis and environmental sensor has the disadvantages of limited storage resources,high computational resource consumption and low precision,a fall detection method in still image based on pose estimation is proposed.The method uses the convolutional neural network(CNN)to extract the pose estimation of the human body,and judges whether the people is a fall state through the pose estimation of the human body,and uses the classification network to verify the fall posture.The experimental results show that the fall detection method in still image based on pose estimation has high recognition rate,low computational resource consumption and fast speed.
作者
杨海清
石珏
YANG Haiqing;SHI Jue(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023,China)
出处
《传感器与微系统》
CSCD
2020年第10期132-134,共3页
Transducer and Microsystem Technologies
关键词
跌倒检测
姿态估计
卷积神经网络
静态图像
fall detection
pose estimation
convolutional neural networks(CNN)
still image
