Graphics Reference
In-Depth Information
and Marsella, 2005). In this model, virtual human movement is divided
into a generalization/specialization hierarchy. Thus, movement is first
performed for the entire body, usually a sitting or standing pose, then
a gesture movement using the arms and torso is performed, then a
separate head and neck movement, then facial and eye movement. By
layering such animations together, it becomes possible to achieve a
large variation in gesture performance. For example, the same gesture
can be combined with several different head or face movements,
producing differing performances. In addition, the hierarchical nature
allows the integration of procedural elements such as gaze control
(Thiebaux et al., 2008) to override specific parts of the body in order
to modify the underlying motion for the specific context in which the
motion in used. The drawback to using such architecture is the loss
of fidelity of the resulting motion; since the synthesized motion was
never originally captured on a human, the dynamics and subtleties
of the synthesized motion will differ from that of a human actor
performing the same motion. Thus, using a hierarchy to generate
gesture performance yields a large variation in performance, at the
expense of motion quality.
3. Multimodal Dynamic: Speech and
Gestures Generation
The generation of multimodal behavior for a virtual human faces
a range of challenges. Most fundamentally is the question of what
behaviors to exhibit. Nonverbal behaviors serve a wide variety of
communicative functions in face-to-face interaction. They can regulate
the interaction: a speaker may avert gaze to hold onto the dialog
turn and may hand off the turn by gazing at a listener. The speaker
can use nonverbal behaviors to convey propositional content: a nod
can convey agreement; raising eyebrows can emphasize a word. The
propositional content of the nonverbal behavior can stand in different
relations to the verbal content, providing information that embellishes,
substitutes for, and even contradicts the information provided verbally.
In other words, the nonverbal behavior is not simply an illustrator
of the verbal information. Nonverbal behaviors also convey a wide
range of mental states and traits: gaze aversions can signal increased
cognitive load, blushing can suggest shyness and facial expressions
can reveal emotional states.
Another challenge here is that this mapping between communicative
function and behaviors is many-to-many. One can emphasize aspects
of the dialog using a hand gesture, a nod or eyebrow raise. On the
Search WWH ::
Custom Search