Joint recognition and segmentation of actions via probabilistic integration of spatio-temporal Fisher vectors

Carvajal, Johanna, McCool, Chris, Lovell, Brian and Sanderson, Conrad (2016). Joint recognition and segmentation of actions via probabilistic integration of spatio-temporal Fisher vectors. In Huiping Cao, Jinyan Li and Ruili Wang (Ed.), Trends and applications in knowledge discovery and data mining (pp. 115-127) Switzerland: Springer. doi:10.1007/978-3-319-42996-0_10

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
carvajal_action_recognition_fisher_vectors.pdf carvajal_action_recognition_fisher_vectors.pdf application/pdf 2.17MB 0

Author Carvajal, Johanna
McCool, Chris
Lovell, Brian
Sanderson, Conrad
Title of chapter Joint recognition and segmentation of actions via probabilistic integration of spatio-temporal Fisher vectors
Title of book Trends and applications in knowledge discovery and data mining
Place of Publication Switzerland
Publisher Springer
Publication Year 2016
Sub-type Research book chapter (original research)
DOI 10.1007/978-3-319-42996-0_10
Open Access Status Not Open Access
Year available 2016
Series Lecture notes in computer science, Lecture notes in artificial intelligence
ISBN 9783319429960
9783319429953
ISSN 1611-3349
0302-9743
Editor Huiping Cao
Jinyan Li
Ruili Wang
Volume number 9794
Chapter number 10
Start page 115
End page 127
Total pages 13
Total chapters 23
Collection year 2017
Language eng
Abstract/Summary We propose a hierarchical approach to multi-action recognition that performs joint classification and segmentation. A given video (containing several consecutive actions) is processed via a sequence of overlapping temporal windows. Each frame in a temporal window is represented through selective low-level spatio-temporal features which efficiently capture relevant local dynamics. Features from each window are represented as a Fisher vector, which captures first and second order statistics. Instead of directly classifying each Fisher vector, it is converted into a vector of class probabilities. The final classification decision for each frame is then obtained by integrating the class probabilities at the frame level, which exploits the overlapping of the temporal windows. Experiments were performed on two datasets: s-KTH (a stitched version of the KTH dataset to simulate multi-actions), and the challenging CMU-MMAC dataset. On s-KTH, the proposed approach achieves an accuracy of 85.0%, significantly outperforming two recent approaches based on GMMs and HMMs which obtained 78.3% and 71.2%, respectively. On CMU-MMAC, the proposed approach achieves an accuracy of 40.9%, outperforming the GMM and HMM approaches which obtained 33.7% and 38.4%, respectively. Furthermore, the proposed system is on average 40 times faster than the GMM based approach.
Keyword Segmentation
Joint recognition
Probabilistic integration
Fisher vectors
Human action recognition
Action recognition
Q-Index Code B1
Q-Index Status Provisional Code
Institutional Status UQ
Additional Notes http://dx.doi.org/10.1007/978-3-319-42996-0_10 http://link.springer.com/chapter/10.1007/978-3-319-42996-0_10

 
Versions
Version Filter Type
Citation counts: Scopus Citation Count Cited 0 times in Scopus Article
Google Scholar Search Google Scholar
Created: Tue, 26 Jul 2016, 01:43:32 EST by Conrad Sanderson on behalf of School of Information Technol and Elec Engineering