Title :
3D Shape Context Based Gesture Analysis Integrated with Tracking using Omni Video Array
Author :
Huang, Kohsia S. ; Trivedi, Mohan M.
Author_Institution :
University of California, San Diego
Abstract :
In this paper we introduce a multilayer cylindrical histogram based feature space for representing 3D shape context of human body. Dynamics of gestures are analyzed using discrete hidden Markov models (DHMM) with quantized feature vectors. Extensive experimental trials with multiple subjects and a range of gesture classes are presented. Gesture recognition accuracies of over 85% (for nine gestures, and 9 subjects) and over 95% (for seven gestures) support the basic feasibility and robustness of the approach.
Keywords :
3D person tracking; Movement analysis; activity analysis; multi-camera systems.; Biological system modeling; Cameras; Computer vision; Context modeling; Hidden Markov models; Humans; Information analysis; Pattern analysis; Robustness; Shape; 3D person tracking; Movement analysis; activity analysis; multi-camera systems.;
Conference_Titel :
Computer Vision and Pattern Recognition - Workshops, 2005. CVPR Workshops. IEEE Computer Society Conference on
Conference_Location :
San Diego, CA, USA
Print_ISBN :
0-7695-2372-2
DOI :
10.1109/CVPR.2005.382