Exploring the Space of an Action for Human Action Recognition

One of the fundamental challenges of recognizing actions is accounting for the variability that arises when arbitrar y cameras capture humans performing actions. In this paper, we explicitly identify three important sources of variabil ity: (1) viewpoint, (2) execution rate, and (3) anthropometry of actors, and propose a model of human actions that allows us to address all three. Our hypothesis is that the variability associated with the execution of an action can be closely approximated by a linear combination of action bases in joint spatio-temporal space. We demonstrate that such a model bounds the rank of a matrix of image measurements and that this bound can be used to achieve recognition of actions based only on imaged data. A test employing principal angles between subspaces that is robust to statistica l fluctuations in measurement data is presented to find the membership of an instance of an action. The algorithm is applied to recognize several actions, and promising result s have been obtained.

[1]  G. Johansson.  Visual perception of biological motion and a model for its analysis , 1973 .

[2]  Karl H.E. Kroemer,et al.  Anthropometry and biomechanics : theory and application , 1982 .

[3]  P. Wedin.  On angles between subspaces of a finite dimensional inner product space , 1983 .

[4]  V. M. Zat︠s︡iorskiĭ.  Kinematics of human motion , 1998 .

[5]  B. Farnell.  Moving Bodies, Acting Selves , 1999 .

[6]  Dariu Gavrila,et al.  The Visual Analysis of Human Movement: A Survey , 1999, Comput. Vis. Image Underst..

[7]  Henning Biermann,et al.  Recovering non-rigid 3D shape from image streams , 2000, Proceedings IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2000 (Cat. No.PR00662).

[8]  Mubarak Shah,et al.  View-invariance in action recognition , 2001, Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001.

[9]  Bernhard P. Wrobel,et al.  Multiple View Geometry in Computer Vision , 2001 .

[10]  M. Alex O. Vasilescu,et al.  Recognizing action events from multiple viewpoints , 2001, Proceedings IEEE Workshop on Detection and Recognition of Events in Video.

[11]  M. Alex O. Vasilescu.  Human motion signatures: analysis, synthesis, recognition , 2002, Object recognition supported by user interaction for service robots.

[12]  Rama Chellappa,et al.  Quasi-invariants for human action representation and recognition , 2002, Object recognition supported by user interaction for service robots.

[13]  Rama Chellappa,et al.  View invariants for human action recognition , 2003, 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings..

[14]  Tanveer F. Syeda-Mahmood,et al.  View-invariant alignment and matching of video sequences , 2003, Proceedings Ninth IEEE International Conference on Computer Vision.

[15]  M. Shah,et al.  On the use of anthropometry in the invariant analysis of human actions , 2004, ICPR 2004.

[16]  Steven M. Seitz,et al.  View-Invariant Analysis of Cyclic Motion , 1997, International Journal of Computer Vision.

[17]  R. Chellappa,et al.  Role of shape and kinematics in human movement analysis , 2004, Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004..

[18]  Takeo Kanade,et al.  Shape and motion from image streams under orthography: a factorization method , 1992, International Journal of Computer Vision.

[19]  Gene H. Golub,et al.  Numerical methods for computing angles between linear subspaces , 1971, Milestones in Matrix Computation.