计算机应用 ›› 2013, Vol. 33 ›› Issue (11): 3187-3189.

• 多媒体处理技术 • 上一篇    下一篇

基于改进朴素贝叶斯分类器的康复训练行为识别方法

张毅1,黄聪1,罗元2   

  1. 1. 重庆邮电大学 信息无障碍工程研发中心,重庆 400065;
    2. 重庆邮电大学 光纤通信技术重点实验室,重庆400065
  • 收稿日期:2013-05-06 修回日期:2013-06-28 出版日期:2013-11-01 发布日期:2013-12-04
  • 通讯作者: 黄聪
  • 作者简介:张毅(1966-),男,重庆潼南人,教授,博士生导师,主要研究方向:机器人、数据融合、信息无障碍技术;黄聪(1990-),女,四川资阳人,硕士研究生,主要研究方向:机器人网络化技术、图像传感与处理;罗元(1972-),女,贵州贵阳人,教授,博士,主要研究方向:机器视觉、智能信号处理、数字图像处理。
  • 基金资助:
    科技部国际合作项目;国家自然科学基金资助项目

Behavior recognition in rehabilitation training based on modified naive bayes classifier

ZHANG Yi1,HUANG Cong1,LUO Yuan2   

  1. 1. Information Accessibility Engineering R&D Center, Chongqing University of Posts and Telecommunications, Chongqing 400065, China;
    2. Key Laboratory of Optical Fiber Communications Technology, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
  • Received:2013-05-06 Revised:2013-06-28 Online:2013-12-04 Published:2013-11-01
  • Contact: HUANG Cong

摘要: 为提高康复训练中行为的识别率,对康复训练行为识别进行研究。首先采用Kinect传感器提取人体骨骼坐标信息,定义运动特征分类集合,完成朴素贝叶斯分类器设计;然后改进康复训练动作识别阈值选择机制提升识别率。改进前后对比实验证明该方法快速简洁,取得了较理想的识别效果。

关键词: 康复训练, Kinect传感器, 阈值选择, 朴素贝叶斯分类器, 行为识别

Abstract: This paper proposed a modified behavior recognition method to improve the recognition rate in rehabilitation training. First, it adopted Kinect sensor to detect human skeleton locations, defined the motion feature in rehabilitation training and designed the Bayes classifier. Second, the threshold selection process was improved to increase the recognition rate. The comparative experimental results with the unmodified one show that the modified naive Bayes classifier is simple and rapid, and it gains better identification effects in rehabilitation training.

Key words: rehabilitation training, Kinect sensor, threshold selection, naive Bayes classifier, behavior recognition

中图分类号: