A Fast Sub-Volume Search Method for Human Action Detection

Ping GUO  Zhenjiang MIAO  Xiao-Ping ZHANG  Zhe WANG  

IEICE TRANSACTIONS on Information and Systems   Vol.E95-D   No.1   pp.285-288
Publication Date: 2012/01/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.E95.D.285
Print ISSN: 0916-8532
Type of Manuscript: LETTER
Category: Image Recognition, Computer Vision
action detection,  graph representation,  Minimum Cycle detection,  sub-volume search,  

Full Text: PDF>>
Buy this Article

This paper discusses the task of human action detection. It requires not only classifying what type the action of interest is, but also finding actions' spatial-temporal locations in a video. The novelty of this paper lies on two significant aspects. One is to introduce a new graph based representation for the search space in a video. The other is to propose a novel sub-volume search method by Minimum Cycle detection. The proposed method has a low computation complexity while maintaining a high action detection accuracy. It is evaluated on two challenging datasets which are captured in cluttered backgrounds. The proposed approach outperforms other state-of-the-art methods in most situations in terms of both Precision-Recall values and running speeds.