Attempts to automate therapist action extraction in robot therapy

Abstract

To make robot therapy more effective, a tool was developed to improve the therapists’ skills by visualizing their knowledge and beliefs. To implement this tool in a busy nursing home, it is necessary to automate a time-consuming process. We developed a system to automate the extraction of the important actions which are recorded by RGBD sensors during the therapy, and conducted an experiment. As the results, 41 ％ of all important actions could be extracted automatically.