计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (24): 164-167.DOI: 10.3778/j.issn.1002-8331.1806-0138

• 模式识别与人工智能 • 上一篇    下一篇

Android平台下的实时人体行为识别

贾小云1,王二虎1,吴敬一2   

  1. 1.陕西科技大学 电气与信息工程学院,西安 710021
    2.东南大学 电子科学与工程学院,南京 210096
  • 出版日期:2018-12-15 发布日期:2018-12-14

Real-time human behavior recognition based on Android platform

JIA Xiaoyun1, WANG Erhu1, WU Jingyi2   

  1. 1.School of Electrical and Information Engineering, Shaanxi University of Science & Technology, Xi’an 710021, China
    2.School of Electronic Science and Engineering, Southeast University, Nanjing 210096, China
  • Online:2018-12-15 Published:2018-12-14

摘要: 为了提高Android平台下实时人体行为识别方法的性能,提出对动作变化和过渡动作进行检测和分割的方法。该方法采用加速度在重力方向上的投影和水平方向上投影的幅值来表征行为活动,通过趋势判断行为变化,结合趋势突变点检测和DTW算法进行过渡动作分割。提取加速度时域特征,使用随机森林对九种行为进行分类识别,平均识别率达到97.26%,其中过渡动作平均识别率达到95.05%。

关键词: 行为识别, 加速度, 过渡动作, 手机传感器

Abstract: In order to improve the performance of the real-time human behavior recognition method based on the Android platform, a method for detecting and segmenting motion changes and transitional motions is proposed. This method uses the projection of the acceleration in the direction of gravity and the amplitude of the projection in the horizontal direction to characterize the behavior. The trend of the acceleration is used to judge the behavior change, and the trend mutation point detection and the DTW algorithm are used to perform the transition action segmentation. Extracting the characteristics of acceleration time domain, nine kinds of behaviors are classified and identified using random forest. The average recognition rate reaches 97.26%, and the average recognition rate of transitional movement reaches 95.05%.

Key words: activity recognition, acceleration, transition motion, smartphone sensors