The performance effects of head tracking and viewpoint correction in a task-based environment

paper a first attempt is made to couple new, cheap head tracking techniques with an existing virtual human framework, resulting in a realistic viewpoint correction mechanism. The research focuses on an environment in which the agent gives users tasks, by pointing at specific objects and locations laid out on a tabletop. The idea is that when a user is able to "look around" the pointing arm of the agent, it is less difficult to identify the correct location from which to pick up or put down an object. During an experiment, using an overhead photo camera the exact locations of selected and placed items are recorded. The results are compared between two groups of participants: one group with, and one without head tracking. Based on a selection of the measured variables there exists a significant accuracy advantage for the head tracking group, in particular the ability to select a specific, correct item from a group of identical items. In a different category of measurements the results are inconclusive, requiring further research on the topic of accurately placing objects once they have been picked up. Keywordstracking, WiiMote, viewpoint correction, task-based environment, virtual human, Elckerlyc, pointing gestures, selection and placement of objects