Human Action Recognition Based On Multi-level Feature Fusion
- 10.2991/cisia-15.2015.96How to use a DOI?
- Human action recognition; multi-level feature; bag-of-words
An efficient multi-level feature fusion descriptor for human action recognition is introduced in the paper. The descriptor is built by the low-level features, which include three trajectory features, HOF and SIFT, combination with the mid-level class correlation feature. Inspired by the recent popularity of dense trajectories in image recognition, they have been utilized to represent actions. It is favorable to extract scene information for action recognition, since human actions have the tightly affinity on specific natural scenes. In addition, noting that different action classes may often share similar motion patterns, we introduce the mid-level class correlation feature to describe relationships among different video classes. Finally, to achieve the better recognition results, bag-of-word model is employed to describe the video by sets of visual words. The average accuracy of the proposed method for action recognition is up to 92.6% on UCF sports dataset.
- © 2015, the Authors. Published by Atlantis Press.
- Open Access
- This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).
Cite this article
TY - CONF AU - Y.Y Xu AU - G.Q Xiao AU - X.Q Tang PY - 2015/06 DA - 2015/06 TI - Human Action Recognition Based On Multi-level Feature Fusion BT - Proceedings of the International Conference on Computer Information Systems and Industrial Applications PB - Atlantis Press SP - 353 EP - 355 SN - 2352-538X UR - https://doi.org/10.2991/cisia-15.2015.96 DO - 10.2991/cisia-15.2015.96 ID - Xu2015/06 ER -