Human Figure Animation
Interacting Modules
• The ones identified– Speech, face, emotion
• Plus others:– Perception– Physiological states
Point of Departure: H-Anim Spec
• Why not view the body as an H-Anim MPEG-4 player?
• To some extent you can– Need to be able to describe body actions as H-Anim
parameters if need be
• But the body usually needs to do more– Execute behaviors
– Perform behaviors autonomously (e.g., breathe)
– Perform simple arbitration between behaviors
Different Bodies with Different Capabilities
• We envision bodies with different degrees of autonomous control– Bodies executing behaviors in fixed ways
– Bodies able to adjust their behavior according to the current emotions of the character
• Goal: use same interface, but behavior depends upon capabilities of the particular virtual human body– Possible mechanism: extensible markup languages
Principle: Full Access to State and Controls
• Body has access to high-level information (e.g., emotion, mood, planned future actions) if it needs it
• No architectural restrictions on information access
• Cognition can choose to control behavior at high level or low level
Example: Gesture Generation
• Body may need to know planned points of emphasis in the speech stream, so that it can time gestures to coincide
• Or defer responsibility to cognition if it is unable to do this
Example: Angry Steve
• Student pulls out dipstick while Steve is demonstrating another step
• Steve needs to decide whether to show anger to the student
• Cognition decides whether to reveal anger to the body
Interaction with Other Modules
• Body and face have parallel access to high level information