中国组织工程研究

中国组织工程研究 ›› 2015, Vol. 19 ›› Issue (53): 8609-8614.doi: 10.3969/j.issn.2095-4344.2015.53.014

• 数字化骨科 digital orthopedics • 上一篇    下一篇

基于随机森林算法的地形识别系统在主动型膝上假肢控制中的应用

张 燕，郜 鑫，陈玲玲，张浩淼   

  1. 河北工业大学控制科学与工程学院，天津市 300130
  • 收稿日期:2015-11-16 出版日期:2015-12-24 发布日期:2015-12-24
  • 通讯作者: 郜鑫，河北工业大学在读硕士，河北工业大学控制科学与工程学院，天津市 300130
  • 作者简介:张燕，女，汉族，1974年生，河北省石家庄市人，2004年南开大学毕业，博士，教授，主要从事智能假肢，非线性系统，智能算法等研究。
  • 基金资助:

     国家自然科学基金项目(61203323，61174009)；河北省高等学校科学技术研究项目(Q2012079)；天津市应用基础与前沿技术研究计划(13JCQNJC03400)

Application of random forest algorithm in terrain recognition to control active above-knee prosthesis

Zhang Yan, Gao Xin, Chen Ling-ling, Zhang Hao-miao   

  1. School of Science and Engineering, Hebei University of Technology, Tianjin 300130, China
  • Received:2015-11-16 Online:2015-12-24 Published:2015-12-24
  • Contact: Gao Xin, Studying for master’s degree, School of Science and Engineering, Hebei University of Technology, Tianjin 300130, China
  • About author:Zhang Yan, M.D., Professor, School of Science and Engineering, Hebei University of Technology, Tianjin 300130, China
  • Supported by:
    the National Natural Science Foundation of China, No. 61203323, 61174009; the Science and Technology Research Project of Hebei Colleges and Universities, No. Q2012079; the Tianjin Application Foundation and Advanced Technology Research Plan, No. 13JCQNJC03400

PDF

1289

被引次数

1

摘要：

背景：在主动型膝上假肢的研究中，现有的运动模式识别方法已经取得了良好的识别效果，但仍需进一步的提高识别精度和缩小响应时间。
目的：建立一个基于随机森林算法的地形识别系统，实现对受试者前方地形的识别，从而获得受试者在步入该地形时的运动模式，应用于假肢的控制当中。
方法：将激光距离传感器和惯性测量单元固定在人体腰部位置，分别采集前方地形信息和人体的运动信息。对采集到的数据进行滤波处理，并提取特征值。利用随机森林算法根据处理过后的数据建立分类器，并进行路况识别。
结果与结论：结果表明，该地形识别系统能够有效的识别出日常行走中常见的平地、上楼梯、下楼梯、上斜坡和下斜坡等5种路况，在主动型膝上假肢的控制中将会发挥重大作用。 

关键词: 骨科植入物, 数字化骨科, 主动型膝上假肢, 运动模式, 识别方法, 随机森林算法, 地形识别系统, 激光距离传感器, 惯性测量单元, 滤波, 特征值, 分类器, 国家自然科学基金

Abstract:

BACKGROUND: In the research of the active above-knee prosthesis, the existing motion pattern recognition methods have shown promising results, but the further improvement of the recognition accuracy and the reduction of the response time are still necessary.
OBJECTIVE: To establish a terrain recognition system based on the random forest algorithm, achieve the identification of the front terrain, and obtain the motion mode of the subject on the terrain for the control of artificial limb.
METHODS: A laser distance sensor and an inertial measurement unit sensor were fixed on the waist to collect the terrain information and human motion signals. The collected data were filtered and the characteristic values were extracted from the data. The random forest algorithm was applied in the establishment of the classifier, which was used to recognize the terrain.
RESULTS AND CONCLUSION: The results showed that the terrain recognition system could recognize level ground, stair ascent/descent and ramp ascent/descent at a high accuracy, which could contribute to the control of the active above-knee prosthesis.  

中图分类号: