Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of GraphSearch.
Virtual Human (VH) creation aims to provide virtual characters with realistic behavior, which implies endowing them with autonomy in an inhabited virtual environment. Autonomous behavior consists in interacting with users or the environment and reacting to stimuli. Reactions are unconscious behaviors that are not often implemented in virtual humans. Frequently, virtual humans show repetitive and robotic movements which tend to decrease realism. To improve believability in virtual humans we need to provide certain level of individuality. This individualization is achieved by using human characteristics like personality, gender, emotions, etc. In this paper, we propose to use those individual descriptors as a basis for different kinds of reactions. Based on observation experiment on real people reacting, we analyze people’s movements and identify different reactions. From this analysis we extract a list of simple movements; and when combining them it is possible to get several different movements. Therefore, we present a semantic diagram (ontology) to classify types of reactions by means of simple movements; and we describe the components involved in the reaction, such as individual descriptors, and virtual human controllers. For our showcase application, we synthesize simple movements using inverse kinematics techniques, and we show that by chaining individual parameters we can produce different reactive animations.