中国安全科学学报 ›› 2019, Vol. 29 ›› Issue (5): 7-12.doi: 10.16265/j.cnki.issn1003-3033.2019.05.002

• 安全人体学 • 上一篇    下一篇

基于BN-SVR的老年人跌倒预测方法

吕子阳1, 王勇毅1 教授, 高星2, 马英楠2 副研究员   

  1. 1 首都经济贸易大学 安全与环境工程学院,北京 100070;
    2 北京城市系统工程研究中心,北京 100035
  • 收稿日期:2019-01-05 修回日期:2019-03-07 发布日期:2020-11-02
  • 作者简介:吕子阳(1995—),男,山西沁源人,硕士研究生,研究方向为安全管理工程、安全人机工程。E-mail:1743945953@qq.com。
  • 基金资助:
    国家重点研发计划基金资助(2018YFC2001400);北京市科学技术研究院创新团队计划课题(IG201704C1);北京城市系统工程研究中心自立课题(2019C010)。

Predicative research on falls in elderly based on BN-SVR

LYU Ziyang1, WANG Yongyi1, GAO Xing2, MA Yingnan2   

  1. 1 School of Safety and Environmental Engineering, Capital University of Economics and Business, Beijing 100070, China;
    2 Beijing Research Center of Urban System Engineering, Beijing 100035, China
  • Received:2019-01-05 Revised:2019-03-07 Published:2020-11-02

摘要: 为减少老年人跌倒事件,以北京市某社区54名老年人为试验对象,开展自由步速下起立行走计时(TUG)的生物力学试验;应用贝叶斯网络(BN)完成运动学数据与跌倒概率的转化,通过支持向量回归(SVR)算法模拟预测姿态轨迹,预测特定帧位的跌倒概率。结果表明:具有显著差异性的髋关节矢状面位移可作为老年人跌倒概率的预测指标,通过对肢体数据的有监督学习,可实现下一时刻老年人跌倒概率的预测,从而可对老年人的高跌倒风险动作进行预警干预。

关键词: 老年人, 贝叶斯网络(BN), 支持向量回归(SVR), 跌倒概率, 起立行走计时(TUG)

Abstract: In order to reduce falls in the elderly, a biomechanical test of TUG at free pace was conducted with 54 elderly people from a community in Beijing; then BN was applied to complete the transformation of kinematics data and fall probability, and the predicted trajectory was simulated with SVR algorithm to predict the fall probability of a specific frame position. The results show that sagittal displacement of the hip joint which features significant difference can be used as a probability predictor for elderly falls; it is also found that through supervised learning of the limb data, predicting the fall probability at the next moment can be realized, thus making it possible to provide preventive intervention for high fall risk actions of the elderly.

Key words: elderly, Bayesian network(BN), support vector regression(SVR), fall probability, timed up and go(TUG)

中图分类号: