Human model and motion based 3D action recognition in multiple view scenarios
Cristian Canton-Ferrer, Josep R. Casas, Montse Pardàs · 2006
This paper presents a novel view-independent approach to the recognition of human gestures of several people in low resolution sequences from multiple calibrated cameras. In contraposition with other multi-ocular gesture recognition systems based on generating a classication on a fusion of features coming from dierent views, our system performs a data fusion (3D representation of the scene) and then a feature extraction and classication. Motion descriptors in-troduced by Bobick et al. for 2D data are extended to 3D and a set of features based on 3D invariant statistical mo-ments are computed. A simple ellipsoid body model is t to incoming 3D data to capture in which body part the gesture occurs thus increasing the recognition ratio of the overall sys-tem and generating a more informative classication output. Finally, a Bayesian classier is employed to perform recogni-tion over a small set of actions. Results are provided showing the eectiveness of the proposed algorithm in a SmartRoom scenario. 1.