A Comparative Evaluation of Action Recognition Methods via Riemannian Manifolds, Fisher Vectors and GMMs: Ideal and Challenging Conditions

Carvajal, Johanna, Wiliem, Arnold, McCool, Chris, Lovell, Brian and Sanderson, Conrad (2016). A Comparative Evaluation of Action Recognition Methods via Riemannian Manifolds, Fisher Vectors and GMMs: Ideal and Challenging Conditions. In: Huiping Cao, Jinyan Li and Ruili Wang, Trends and Applications in Knowledge Discovery and Data Mining. Pacific-Asia Conference on Knowledge Discovery and Data Mining, Auckland, New Zealand, (88-100). 19-22 April 2016. doi:10.1007/978-3-319-42996-0_8

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads

Author Carvajal, Johanna
Wiliem, Arnold
McCool, Chris
Lovell, Brian
Sanderson, Conrad
Title of paper A Comparative Evaluation of Action Recognition Methods via Riemannian Manifolds, Fisher Vectors and GMMs: Ideal and Challenging Conditions
Conference name Pacific-Asia Conference on Knowledge Discovery and Data Mining
Conference location Auckland, New Zealand
Conference dates 19-22 April 2016
Proceedings title Trends and Applications in Knowledge Discovery and Data Mining   Check publisher's open access policy
Journal name Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)   Check publisher's open access policy
Series Lecture Notes in Computer Science
Publisher Springer
Publication Year 2016
Year available 2016
Sub-type Fully published paper
DOI 10.1007/978-3-319-42996-0_8
Open Access Status Not yet assessed
ISBN 9783319429953
9783319429960
ISSN 1611-3349
0302-9743
Editor Huiping Cao
Jinyan Li
Ruili Wang
Volume 9794
Start page 88
End page 100
Total pages 13
Language eng
Abstract/Summary We present a comparative evaluation of various techniques for action recognition while keeping as many variables as possible controlled. We employ two categories of Riemannian manifolds: symmetric positive definite matrices and linear subspaces. For both categories we use their corresponding nearest neighbour classifiers, kernels, and recent kernelised sparse representations. We compare against traditional action recognition techniques based on Gaussian mixture models and Fisher vectors (FVs). We evaluate these action recognition techniques under ideal conditions, as well as their sensitivity in more challenging conditions (variations in scale and translation). Despite recent advancements for handling manifolds, manifold based techniques obtain the lowest performance and their kernel representations are more unstable in the presence of challenging conditions. The FV approach obtains the highest accuracy under ideal conditions. Moreover, FV best deals with moderate scale and translation changes.
Subjects 2614 Theoretical Computer Science
1700 Computer Science
Keyword Computer Science, Artificial Intelligence
Computer Science, Interdisciplinary Applications
Computer Science
Q-Index Code E1
Q-Index Status Provisional Code
Institutional Status UQ

Document type: Conference Paper
Sub-type: Fully published paper
Collections: HERDC Pre-Audit
School of Information Technology and Electrical Engineering Publications
 
Versions
Version Filter Type
Citation counts: TR Web of Science Citation Count  Cited 1 times in Thomson Reuters Web of Science Article | Citations
Scopus Citation Count Cited 0 times in Scopus Article
Google Scholar Search Google Scholar
Created: Sat, 28 May 2016, 00:41:28 EST by Arnold Wiliem on behalf of School of Information Technol and Elec Engineering