Action Capture: A VR-Based Method for Character Animation

This contribution describes a Virtual Reality (VR) based method for character animation that extends conventional motion capture by not only tracking an actor’s movements but also his or her interactions with the objects of a virtual environment. Rather than merely replaying the actor’s movements, the idea is that virtual characters learn to imitate the actor’s goal-directed behavior while interacting with the virtual scene. Following Arbib’s equation action = movement + goal we call this approach Action Capture. For this, the VR user’s body movements are analyzed and transformed into a multi-layered action representation. Behavioral animation techniques are then applied to synthesize animations which closely resemble the demonstrated action sequences. As an advantage, captured actions can often be naturally applied to virtual characters of different sizes and body proportions, thus avoiding retargetting problems of motion capture.