On the timing of gestures of a Virtual Physiotherapist

A Reactive Virtual Trainer is a virtual human capable to demonstrate physical exercises, to monitor the perform-ance of the user and react in speech, both addressing the correctness of the motion, the overall performance and the motivation of the user. In this paper we focus on the timing of the simple exercises motions (called gestures). We discuss the subtle requirements on synchronizing motion to speech and other acoustic modalities, and the proper timing of the stages of the gesture. We outline our implemented demo, to visualize and to experiment with the different planning and timing strategies. Finally, we talk about further work concerning motion models and real-time adaptive planning for our RVT application.