.png){kind=icon}
.png){kind=icon}
Return
Motion pattern based representation for improving human action retrieval
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-018-5835-6.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In recent years, many videos have been shared on the Internet. Finding desired videos among the large amount of video files is known as a major challenge in big data. Due to various problems in tag based search methods, content based search has been introduced as an alternative way for finding desired videos. On the other hand, there are many relative videos to humans and human actions are considered as the basis content for searching videos that is called human action retrieval. Action retrieval methods use some features (whether global or local) to represent the actions. Due to the stability of local features against noise, local features are used for representing the human action. In this paper, we represent the human actions by the pattern of body motions. Focusing on patterns can capture more accurate information about the actions. To this end, the complexity of the pattern of the body motions on different scales of coordinate axes is calculated to describe the body motion and the human action. The complexity of motion patterns is calculated by inspiration from the fractal dimension calculation. Our method can discriminate different actions, which have similar movements, more accurate than the state-of-the-art methods. Experimental results on KTH, UCFYT, and HMDB datasets shows the better performance of our method than the state-of-the-art methods.
Keywords:
Content Based Video Retrieval
Big data
Human action
STIP
Motion pattern
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
3
Papers:
1.9W
Citations:
3.2W