Accurate 3D action recognition using learning on the Grassmann manifold

被引:150
作者
Slama, Rim [1 ,2 ]
Wannous, Hazem [1 ,2 ]
Daoudi, Mohamed [2 ,3 ]
Srivastava, Anuj [4 ]
机构
[1] Univ Lille 1, F-59655 Villeneuve Dascq, France
[2] CNRS, UMR 8022, LIFL Lab, Villeneuve Dascq, France
[3] Inst Mines Telecom Telecom Lille, Villeneuve Dascq, France
[4] Florida State Univ, Dept Stat, Tallahassee, FL 32306 USA
基金
美国国家科学基金会;
关键词
Human action recognition; Grassmann manifold; Observational latency; Depth images; Skeleton; Classification; SPARSE REPRESENTATION; VIDEO; ALGORITHMS;
D O I
10.1016/j.patcog.2014.08.011
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In this paper we address the problem of modeling and analyzing human motion by focusing on 3D body skeletons. Particularly, our intent is to represent skeletal motion in a geometric and efficient way, leading to an accurate action-recognition system. Here an action is represented by a dynamical system whose observability matrix is characterized as an element of a Grassmann manifold. To formulate our learning algorithm, we propose two distinct ideas: (1) in the first one we perform classification using a Truncated Wrapped Gaussian model, one for each class in its own tangent space. (2) In the second one we propose a novel learning algorithm that uses a vector representation formed by concatenating local coordinates in tangent spaces associated with different classes and training a linear SVM. We evaluate our approaches on three public 3D action datasets: MSR-action 3D, UT-kinect and UCF-kinect datasets; these datasets represent different kinds of challenges and together help provide an exhaustive evaluation. The results show that our approaches either match or exceed state-of-the-art performance reaching 91.21% on MSR-action 3D, 97.91% on UCF-kinect, and 88.5% on UT-kinect. Finally, we evaluate the latency, i.e. the ability to recognize an action before its termination, of our approach and demonstrate improvements relative to other published approaches. (C)2014 Elsevier Ltd. All rights reserved.
引用
收藏
页码:556 / 567
页数:12
相关论文
共 56 条
[1]   Silhouette-based gesture and action recognition via modeling trajectories on Riemannian shape manifolds [J].
Abdelkader, Mohamed F. ;
Abd-Almageed, Wael ;
Srivastava, Anuj ;
Chellappa, Rama .
COMPUTER VISION AND IMAGE UNDERSTANDING, 2011, 115 (03) :439-455
[2]   Human Activity Analysis: A Review [J].
Aggarwal, J. K. ;
Ryoo, M. S. .
ACM COMPUTING SURVEYS, 2011, 43 (03)
[3]   Human activity recognition using multi-features and multiple kernel learning [J].
Althloothi, Salah ;
Mahoor, Mohammad H. ;
Zhang, Xiao ;
Voyles, Richard M. .
PATTERN RECOGNITION, 2014, 47 (05) :1800-1812
[4]  
[Anonymous], IEEE T PATTERN ANAL
[5]  
[Anonymous], MACHINE LEARNING COM, DOI [DOI 10.1145/2398356.2398381, 10.1145/2398356.2398381]
[6]  
[Anonymous], 2011, Acm T. Intel. Syst. Tec., DOI [DOI 10.1145/1961189.1961199, 10. 1145/1961189.1961199]
[7]   Grassmannian Sparse Representations and Motion Depth Surfaces for 3D Action Recognition [J].
Azary, Sherif ;
Savakis, Andreas .
2013 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS (CVPRW), 2013, :492-499
[8]  
Azary S, 2012, IEEE IMAGE PROC, P769, DOI 10.1109/ICIP.2012.6466973
[9]   Ongoing human action recognition with motion capture [J].
Barnachon, Mathieu ;
Bouakaz, Saida ;
Boufama, Boubakeur ;
Guillou, Erwan .
PATTERN RECOGNITION, 2014, 47 (01) :238-247
[10]  
Bautista Miguel Angel, 2013, Advances in Depth Image Analysis and Applications. International Workshop, WDIA 2012. Selected and Invited Papers: LNCS 7854, P126, DOI 10.1007/978-3-642-40303-3_14