Journal of South China University of Technology(Natural Science Edition) ›› 2012, Vol. 40 ›› Issue (1): 146-151.

• Computer Science & Technology • Previous Articles     Next Articles

Human Action Recognition in Complex Scenes Based on Fuzzy Integral Fusion

Wu Qiu-xia  Deng Fei-qi  Kang Wen-xiong   

  1. School of Automation Science and Engineering,South China University of Technology,Guangzhou 510640,Guangdong,China
  • Received:2011-06-07 Revised:2011-09-27 Online:2012-01-25 Published:2011-12-01
  • Contact: 康文雄(1976-) ，男，博士，讲师，主要从事计算机视觉及生物特征识别研究．E-mail:auwxkang@scut.edu.cn E-mail:wutong_924@163.com
  • About author:吴秋霞(1983-) ，女，博士生，主要从事计算机视觉方面的研究．
  • Supported by:

    国家自然科学基金资助项目( 60874114， 61105019) ; 广东省自然科学基金资助项目( S2011040002474) ; 广东省科技计划项目( 2011B010200023) ; 广东省教育部产学研结合项目( 2011B090400564) ; 华南理工大学中央高校基本科研业务费专项资金资助项目( 2012ZZ0108)

PDF

1153

Abstract:

The approach to representing a series of key points in videos by using local feature descriptors has been widely applied to the recognition of human action in complex scenes. However,the important structural information among the key points has not been investigated yet. In this paper,first,a scale-invariant key point detector and a 3D-Harris detector are used to find the local key points in video samples. Next,the existing local feature descriptor and shape descriptor are employed to describe the structural information about the positions of the key points. Then,the bag-of-features model is utilized to calculate the distribution of the features. Finally,the fuzzy integral scheme is used to fuse the local features,with the corresponding algorithm being also described. It is found form the experiments on the YouTube dataset in complex scenes that the proposed approach to local feature description effectively represents the human action in complex scenes,and that the fuzzy integral fusion scheme is effective in integrating the advantages of the descriptors on the decision level.

Key words: human action recognition, shape descriptor, local feature, fuzzy integral fusion