Behavioral Animation - Computer Animation: Algorithms and Techniques - page 357

and disgust. Ekman [ 12 ] has also suggested there are 15 emotions: amusement, anger, contempt, con-
tentment, disgust, embarrassment, excitement, fear, guilt, pride in achievement, relief, sadness/distress,
satisfaction, sensory pleasure, and shame. The Ortony Clore Collins (OCC) model describes how the
perceptions of people dictate how they experience emotions [ 28 ] .
The Improv system [ 29 ] provides tools for an animator to script the behavior of characters that
respond to users and to other characters in real time. It consists of a behavior engine that is responsible
for executing, in parallel, scripts that are applicable to the current situation. The scripts allow for lay-
ered behavior so that the animator can build high-level behaviors on top of low-level ones. It provides
for nondeterministic behavior by including probabilistic weights to be associated with alternative
choices. Underneath the behavior engine is the animation engine, which is responsible for procedurally
generating motion segments and blending between them. Personality and emotions are not explicitly
modeled but must be encoded in the choices and probabilities supplied with the scripted behavior.
The AlphaWolf project of Tomlinson and Blumberg [ 40 ] models three parameters of emotion: plea-
sure, arousal, and dominance. These are updated based on past emotion, an emotion's rate of drift, and
environmental factors. In addition, there is context-specific emotional memory that can bring a char-
acter back to a previously experienced emotion.
Egges et al. described a system based on the OCC model [ 8 ][ 9 ]. The big five personality traits are
used to modify emotional states. An intelligent three-dimensional talking head with lip-sync speech
and emotional states creates behavior. Emotions are used to control the conversational behavior of
an agent. Amethod is employed to update a character's emotional state based on personality, emotions,
mood, and emotional influences.
11.4 Crowds
A crowd is a large group of people physically grouped (crowded) together. But the term “crowd” has
been applied to a wide variety of scenarios, such as a collection of individuals reacting to the environ-
ment (e.g., [ 42 ] ), a mass of hundreds of thousands of armed warriors in a feature length film where
individual identification is practically impossible, and dots on a screen showing movement from room
to room during a fire simulation. For this discussion, a crowd is considered to be the animation of rep-
resentations of a large number of individuals where the individuals are not the primary focus of the
animation and detailed inspection of individuals by the viewer is not expected. This distinguishes a
crowd from an environment containing multiple agents in which the individual activity of the agents
is important. The use of multiple agents is just that—some number of behaviorally animated characters
where the actions of individuals are discernible and important and the viewer would be inspecting the
activity of individuals in the environment. The emphasis here is on animation of multiple figures where
the global nature of the activity is the primary consideration.
There are two main applications of graphical crowd modeling. The first is as a visual effect. Perhaps
the obvious one is the use in feature length film to fill in the movement or activity of a mass of people in
battle scenes or populating an environment with activity or creating an audience in the bleachers of a
stadium. In battle scenes, for example, it is common to have the first row or two of characters to be live
actors while all the background activity is synthetic characters. Managing the complexity of such
crowds has become an important issue in the visual effects industry and is being addressed by software
products such as Massive [ 20 ] [ 23 ] . Computer games are another obvious example of this application.