计算机科学 ›› 2019, Vol. 46 ›› Issue (9): 106-112.doi: 10.11896/j.issn.1002-137X.2019.09.014

• 第35届中国数据库学术会议 • 上一篇    下一篇

基于深度学习的跌倒行为识别

马露1, 裴伟2, 朱永英3, 王春立1, 王鹏乾1   

  1. (大连海事大学信息科学技术学院 辽宁 大连116026)1;
    (大连海事大学环境科学与工程学院 辽宁 大连116026)2;
    (大连海洋大学海洋与土木工程学院 辽宁 大连116026)3
  • 收稿日期:2018-07-09 出版日期:2019-09-15 发布日期:2019-09-02
  • 通讯作者: 裴 伟(1977-),男,博士,副教授,主要研究方向为图像处理、数据挖掘,E-mail:peiwei2002@163.com
  • 作者简介:马 露(1993-),女,硕士生,主要研究方向为图像处理、异常行为识别,E-mail:malu930310@163.com;朱永英(1978-),女,博士,副教授,主要研究方向为模式识别;王春立(1972-),女,教授,主要研究方向为模式识别与数据挖掘;王鹏乾(1997-),男,主要研究方向为电子信息工程、数据挖掘。
  • 基金资助:
    国家自然科学基金项目(61001158,61001158,61370070),辽宁省自然科学基金项目(2014025003),辽宁省教育厅科学研究一般项目(L2012270),大连市科技创新基金(2018J12GX043),辽宁省重点研发计划指导计划项目

Fall Action Recognition Based on Deep Learning

MA Lu1, PEI Wei2, ZHU Yong-ying3, WANG Chun-li1, WANG Peng-qian1   

  1. (College of Information Science and Technology,Dalian Maritime University,Dalian,Liaoning 116026,China)1;
    (College of Environmental Science and Engineering,Dalian Maritime University,Dalian,Liaoning 116026,China)2;
    (Ocean and Civil Engineering Department,Dalian Ocean University,Dalian,Liaoning 116026,China)3
  • Received:2018-07-09 Online:2019-09-15 Published:2019-09-02

摘要: 随着老龄人口的快速增长,跌倒检测成为医疗健康领域的一个关键问题。准确检测监控视频中的跌倒行为并及时反馈能有效减少老年人因跌倒造成的伤害甚至死亡。针对监控视频中的复杂场景及多种相似性人类行为干扰的情况,文中提出一种改进的FSSD(Feature Fusion Single Shot Multibox Detector)跌倒检测方法。首先,从不同的跌倒视频序列中抽取视频帧形成数据集;然后,将训练样本集输入到改进的FSSD网络中训练直至网络收敛;最后,根据最优化的网络模型测试视频中目标的类别并定位目标。实验结果表明,改进的FSSD 算法可以有效检测每帧图像的跌倒或日常生活活动(Activities of Daily Living,ADL)事件并给出实时反馈,检测速度为24fps(GTX1050Ti),在保证检测精度的同时满足实时性要求。将改进方法与已有最新方法进行比较,结果表明:改进的FSSD 算法的性能优于其他算法。视频中跌倒行为的检测进一步验证了基于深度学习的识别方法的可行性与高效性。

关键词: 跌倒检测, 卷积神经网路, FSSD目标检测算法, 深度学习, 行为检测

Abstract: With the rapid growth of the aging population,fall detection has become a key issue in the medical and health field.Accurately detecting falling events in the monitoring video and giving feedback in real time can effectively reduce injuries even deaths caused by falls in the elderly.In view of the complex scenes in the monitoring video and multiple similar human behaviors,this paper proposed an improved FSSD (Feature Fusion Single Shot Multibox Detector) fall detection method.Firstly,a video frame forming dataset is extracted from different falling video sequences.Then,the training sample set is input into the improved convolutional neural network until the network converges.Finally,the target category and the location of the target in the video are tested according to the optimized network model.The experimental results show that the improved FSSD algorithm can effectively detect the falling or ADL activities of each frame of image and provide real-time feedback.The detection speed is 24fps (GTX1050Ti),which can meet the real-time requirements while ensuring the detection accuracy.Comparing the improved method with the state-of-the-art fall detection methods,the performance of the improved FSSD is better than other algorithms.The detection of fall behavior in video further validates the feasibility and efficiency of the recognition method based on deep learning.

Key words: Fall detection, Convolutional neural network, FSSD target detection algorithm, Deep learning, Action detection

中图分类号: 

  • TP391
[1]PENG Y A.A Technical Guide to Falling Interventions for the Elderly[J].Journal of Practical Rural Doctors,2012,19(8):1-13.(in Chinese)彭宇案.老年人跌倒干预技术指南[J].中国实用乡村医生杂志,2012,19(8):1-13.
[2]SANNINO G,FALCO I DE,PIETRO G DE.A supervised approach to automatically extract a set of rules to support fall detection in an mHealth system[J].Applied Soft Computing,2015,34(C):205-216.
[3]National Bureau of Statistics of the People’s Republic of China. Statistical Communique of the 2017 National Economic and Social Development of the People’s Republic of China[J].China Statistics,2018(3):7-20.(in Chinese)中国人民共和国国家统计局.中华人民共和国2017年国民经济和社会发展统计公报[J].中国统计,2018(3):7-20.
[4]NOURY N,FLEURY A,RUMEAU P,et al.Fall Detection-principles and Methods//29th Annual International Conference of the IEEE on Engineering in Medicine and Biology Society.Lyon,France,2007:1663-1666.
[5]MUBASHIR M,SHAO L,SEED L.A survey on fall detection:Principles and approaches[J].Neurocomputing,2013,100:144-152.
[6]RATHI N,KAKANI M,EL-SHARKAWY M,et al.Wearablelow power pre-fall detection system with IoT and bluetooth capabilities[C]//IEEE National Aerospace and Electronics Conference (NAECON).Dayton,OH,2017:241-244.
[7]HOSSAIN F,ALI M L,ISLAM M Z,et al.A direction-sensitive fall detection system using single 3D accelerometer and learning classifier[C]//International Conference on Medical Engineering,Health Informatics and Technology (MediTec).Dhaka,2016:1-6.
[8]SCHWICKERT L,BECKER C,LINDEMANN U,et al.Fall detection with body-worn sensors?:a systematic review[J].Zeitschrift Für Gerontologie Und Geriatrie,2013,46(8):706-719.
[9]YAZAR A,ÇETIN A E.Ambient assisted smart home designusing vibration and PIR sensors[C]//21st Signal Processing and Communications Applications Conference (SIU).Haspolat,2013:1-4.
[10]ARSHAD A,KHAN S,ALAM A H M Z,et al.A capacitive proximity sensing scheme for human motion detection[C]//IEEE International Instrumentation and Measurement Techno-logy Conference(I2MTC).Turin,2017:1-5.
[11]REN S,GIRSHICK R,GIRSHICK R,et al.Faster R-CNN:Towards Real-Time Object Detection with Region Proposal Networks[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2015,39(6):1137-1149.
[12]REDMON J,DIVVALA S,GIRSHICK R,et al.You Only Look Once:Unified,Real-Time Object Detection[C]//IEEE Confer-ence on Computer Vision and Pattern Recognition (CVPR).Las Vegas,NV,2016:779-788.
[13]LIU W,ANGUELOV D,ERHAN D,et al.SSD:Single shotmultibox detector[J].arXiv:1512.02325v2,2015.
[14]ZHOU X Y,WANG K,LI L Y.Review of object detection based on deep learning[J].Electronic Measurement Technology,2017(11):89-93.(in Chinese)周晓彦,王珂,李凌燕.基于深度学习的目标检测算法综述[J].电子测量技术,2017(11):89-93.
[15]FU C,LIU W,RANGA A,et al.DSSD:Deconvolutional single shot detector[J].arXiv:1701.06659,2016.
[16]JEONG J,PARK H,KWAK N.Enhancement of ssd by concatenating feature maps for object detection[J].arXiv:1705.09587,2017.
[17]LI Z,ZHOU F.FSSD:Feature fusion single shot multibox detector[J].arXiv preprint arXiv:1712.00960,2017.
[18]SHEN Z,LIU Z,LI J,et al.DSOD:Learning Deeply Supervised Object Detectors from Scratch//Proceedings of the IEEE International Conference on Computer Vision.2017:1937-1945.
[19]ZHANG Z,CONLY C,ATHITSOS V.A survey on vision-based fall detection[C]//8th ACM International Conference on PErvasive Technologies Related to Assistive Environments.2015:1-7.
[20]MITERAN J,DUBOIS J,ATRI M.Optimized spatio-temporal descriptors for real-time fall detection:comparison of support vector machine and Adaboost-based classification[J].Journal of Electronic Imaging,2013,22(4):41106.
[21]YUN Y,INNOCENTI C,NERO G,et al.Fall detection in RGB-D videos for elderly care[C]//17th International Conference on E-health Networking.Boston,MA,2015:422-427.
[22]LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436.
[23]ADHIKARI K,BOUCHACHIA H,NAIT-CHARIF H.Activity recognition for indoor fall detection using convolutional neural network[C]//Fifteenth IAPR International Conference on Machine Vision Applications.Nagoya,2017:81-84.
[24]HUANG C D,WANG C Y,WANG J.Human action recognition system for elderly and children care using three stream ConvNet[C]//International Conference on Orange Technologies.Hong Kong,2016:5-9.
[25]SIMONYAN K,ZISSERMAN A.Two-stream convolutionalnetworks for action recognition in videos[J].Advances in Neural Information Processing Systems,2014,1(4):568-576.
[26]MIN W,CUI H,RAO H,et al.Detection of Human Falls on Furniture Using Scene Analysis Based on Deep Learning and Activity Characteristics[J].IEEE Access,2018,PP(99):1.
[27]HE K,SUN J.Convolutional neural networks at constrainedtime cost[C]//IEEE Conference on Computer Vision and Pattern Recognition (CVPR).2015:5353-5360.
[28]ROUGIER C,MEUNIER J,ST-ARNAUD A,et al.Robust Video Surveillance for Fall Detection Based on Human Shape Deformation.IEEE Transactions on Circuits and Systems for Video Technology,2011,21(5):611-622.
[29]MIRMAHBOUB B,SAMAVI S,KARIMI N,et al.Automatic monocular system for human fall detection based on variations in silhouette area[J].IEEE Transactions on Biomedical Engineering,2013,60(2):427-436.
[1] 李青华, 李翠平, 张静, 陈红, 王绍卿. 深度神经网络压缩综述[J]. 计算机科学, 2019, 46(9): 1-14.
[2] 王嫣然, 陈清亮, 吴俊君. 面向复杂环境的图像语义分割方法综述[J]. 计算机科学, 2019, 46(9): 36-46.
[3] 孙中锋, 王静. 用于基于方面情感分析的RCNN-BGRU-HN网络模型[J]. 计算机科学, 2019, 46(9): 223-228.
[4] 缪永伟, 李高怡, 鲍陈, 张旭东, 彭思龙. 基于卷积神经网络的图像局部风格迁移[J]. 计算机科学, 2019, 46(9): 259-264.
[5] 周燕, 曾凡智, 吴臣, 罗粤, 刘紫琴. 基于深度学习的三维形状特征提取方法[J]. 计算机科学, 2019, 46(9): 47-58.
[6] 郭旭, 朱敬华. 基于用户向量化表示和注意力机制的深度神经网络推荐模型[J]. 计算机科学, 2019, 46(8): 111-115.
[7] 张义杰, 李培峰, 朱巧明. 基于自注意力机制的事件时序关系分类方法[J]. 计算机科学, 2019, 46(8): 244-248.
[8] 邓存彬, 虞慧群, 范贵生. 融合动态协同过滤和深度学习的推荐算法[J]. 计算机科学, 2019, 46(8): 28-34.
[9] 杜威, 丁世飞. 多智能体强化学习综述[J]. 计算机科学, 2019, 46(8): 1-8.
[10] 李舟军,王昌宝. 基于深度学习的机器阅读理解综述[J]. 计算机科学, 2019, 46(7): 7-12.
[11] 张琳娜,陈建强,陈晓玲,岑翼刚,阚世超. 面向行车视频目标实时检测的轻量级SSD网络[J]. 计算机科学, 2019, 46(7): 233-237.
[12] 李健, 杨祥如, 何斌. 基于深度学习的几何特征匹配方法[J]. 计算机科学, 2019, 46(7): 274-279.
[13] 刘梦娟,曾贵川,岳威,仇笠舟,王加昌. 面向展示广告的点击率预测模型综述[J]. 计算机科学, 2019, 46(7): 38-49.
[14] 陈思文, 刘玉江, 刘冬, 苏晨, 赵地, 钱林学, 张佩珩. 基于AlexNet模型和自适应对比度增强的乳腺结节超声图像分类[J]. 计算机科学, 2019, 46(6A): 146-152.
[15] 霍星, 费志伟, 赵峰, 邵堃. 深度学习在驾驶员安全带检测中的应用[J]. 计算机科学, 2019, 46(6A): 182-187.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
[1] 梁俊斌,周翔,王田,李陶深. 移动低占空比无线传感网中数据收集的研究进展[J]. 计算机科学, 2018, 45(4): 19 -24, 52 .
[2] 张昱, 高克宁, 于戈. 一种融合节点属性信息的社会网络链接预测方法[J]. 计算机科学, 2018, 45(6): 41 -45 .
[3] 李航, 臧洌, 甘露. 基于蚁群算法的猜测符号执行的路径搜索[J]. 计算机科学, 2018, 45(6): 145 -150 .
[4] 吕巨建, 赵慧民, 陈荣军, 李键红. 基于自适应稀疏邻域重构的无监督主动学习算法[J]. 计算机科学, 2018, 45(6): 251 -258 .
[5] 屠要峰,刘辉,张国良,刘春. 一种分布式缓存系统的关键技术及应用[J]. 计算机科学, 2018, 45(5): 156 -162 .
[6] 赵星宇, 丁世飞. 深度强化学习研究综述[J]. 计算机科学, 2018, 45(7): 1 -6 .
[7] 刘壮,柴秀娟,陈熙霖. 双通道Faster R-CNN在RGB-D手部检测中的应用[J]. 计算机科学, 2018, 45(5): 232 -237 .
[8] 李小薪,吴克宋,齐盼盼,周旋,刘志勇. 局部球面规范化嵌入:PCANet的一种改进方案[J]. 计算机科学, 2018, 45(5): 238 -242, 249 .
[9] 刘春阳, 吴泽民, 胡磊, 刘熹. 基于似物性和空时协方差特征的行人检测算法[J]. 计算机科学, 2018, 45(6A): 210 -214, 246 .
[10] 王振朝,赵云,薛文玲. 下含D2D蜂窝网中基于公平性原理的功率控制[J]. 计算机科学, 2018, 45(7): 104 -109 .