We used models of human movements (which we labeled primitives) to build distributions of movement corresponding to the individual models. A Bayesian classifier was then implemented to categorize observed movement. This method performs very well, and is described in the following two papers:
- Evan Drumwright, Odest Jenkins, and Maja Mataric'. Exemplar-Based Primitives for Humanoid Movement Classification and Control. In IEEE Intl. Conf. on Robotics and Automation (ICRA), pp. 140-145, Apr 2004.
- Evan Drumwright and Maja Mataric'. Generating and Recognizing Free-Space Movement in Humanoid Robotics. In IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems (IROS), pp. 1672-1678, Las Vegas, NV, Oct 2003.