Therapeutic interaction detection for serious games in physical rehabilitation

Serious games gained popularity in recent years together with the use of modern input devices. Mainly marker-less motion tracking cameras play a special role in the automation of physical rehabilitation. These inexpensive cameras can provide accurate information about the movements and poses of the subject without complicated setup. However, these cameras are still not perfect and experience problems in particular poses, setups or when users are interacting. Interaction between a patient and the therapist is a crucial and inevitable aspect of the therapy and results in frustrations when using new technologies. In this paper we propose a method that can identify whether a therapist is interacting with a patient or not, in order to improve not only the therapy sessions but also the quality of the data collected during the gameplay or assessment, automated with the modern input sensors. We compare our measurement results with a marker based motion tracking system (Vicon) and additional scores to demonstrate the importance of identifying interactions between a therapist and patients.