计算机与现代化

• 计算机仿真 • 上一篇    下一篇

基于神经网络的人体动作识别方法

  

  1. (合肥工业大学宣城校区信息工程系,安徽宣城242000)
  • 收稿日期:2017-07-12 出版日期:2018-04-03 发布日期:2018-04-03
  • 作者简介:董哲宇(1995-),男,河北张家口人,合肥工业大学宣城校区信息工程系本科生,研究方向:计算机视觉; 汪千军(1995-),男,安徽安庆人,本科生,研究方向:计算机视觉; 李万杰(1995-),男,重庆人,本科生,研究方向:计算机视觉; 通信作者:周波(1981-),男,湖北宜昌人,讲师,博士,研究方向:地理信息系统。
  • 基金资助:
    国家自然科学基金资助项目(41401445); 安徽省大学生创新创业训练基金资助项目(2016CXCYS111)

Human Activity Recognition Method Based on Neural Network

  1. (Information Engineering Department, Xuancheng Campus, Hefei University of Technology, Xuancheng 242000, China)
  • Received:2017-07-12 Online:2018-04-03 Published:2018-04-03

摘要: 人体动作识别一直是计算机视觉领域的研究重点。为了提高人体动作识别的准确度,本文提出一种基于神经网络的加权识别方法。首先利用ViBe算法提取人体运动前景,计算前景重心,然后将轮廓重心距作傅里叶变换获得傅里叶描述子,最后利用本文提出的基于神经网络的加权识别方法进行分类。实验结果表明,本文方法的识别率在89%以上。

关键词: 动作识别, 神经网络, 傅里叶描述子, ViBe, 加权识别

Abstract: Human activity recognition has always been paid attention to the field of computer vision. In this paper, a weighted recognition method based on neural network is presented to improve the accuracy of human activity recognition. Firstly, the ViBe algorithm is used to extract the foreground of human activity, and the center of gravity of the foreground is calculated. Secondly, the Fourier descriptor is obtained by the Fourier transform of the outline distance center of gravity. Finally, a weighted recognition method based on neural network is used to classify the Fourier descriptor. The experimental results show that the recognition rate of this method is more than 89%.

Key words: activity recognition, neural network, Fourier descriptor, ViBe, weighted recognition

中图分类号: