Braun, et al.
Norbert Braun and Thomas Rieger
{nbraun,rieger}@gris.informatik.tu-darmstadt.de
Expressiveness Generation for Virtual Characters in Games
Extended Abstract
While a conversation, a human unknowingly expects some actions and reactions from a conversational partner in a discourse, likewise from a Virtual Character (VC). These actions and reactions can be verbal and/or nonverbal. The nonverbal actions and reactions are expressed with facial expression, gaze, body position, gesture, etc., they are content and context dependent:
ð Humans use story-adapted nonverbal expressions to narrate stories to each other (content
dependence).
ð They use di_erent sets of nonverbal actions when changing the social context of the conversation, e.g. while narrating things to a superior or to a relative.
The same should hold for the facial expression and the gesture of a Virtual Character used in Games. The VC should adapt its expressiveness to the content and the context of the human-computerinteraction discourse.To achieve this, we use the output of an actoraction- based story generation module, a so called Story Engine, see Braun [1, 2], to adapt the suspense-level of the content narration: The actor-action-based approach is based on a semiotic approach of Russian formalist Propp [3] to describe conflict potential within a story: Propp describes a story (he used Russian Folk (or Fairy) Tales for his analysis) as an abstract set of morphological functions. Every function has some action content and some abstract characters playing this content. The functions are connected to each other via inter-functional dependencies, which allows Propp to state rules that are usable to process the functions to a coherent story. This approach comprised some important implicit information within the several functions - the information about the start of a conflict, the ongoing conflict, the chance to solve the conflict and the resolution, as well as the prize to be attained upon solving the conflict. We call this information the Narrative Conflict. This needs to be combined with a suspense model based on a time model of suspense, like the Aristotle model (he o_ered a five-act structure (exposition, rising action, retarding (crisis),climax, denouement). We call this time model the Suspense Progression. The two concepts of Narrative Conflict and Suspense Progression are used to extend a Speech Acts approach. Using this information, the expressiveness of the Virtual Character can be modified by using Narrative Extended Speech Acts (NESA), built up from Speech Acts and the narrative meta information. The Narrative Conflict and Suspense Progression values are calculated to an expressiveness value within the interval [0..1], used to control the expressiveness of a Virtual Character. The figures 1 and 2 show the discrepancy in the gesture and facial expression expressiveness, when using different narrative information.
Biography
Norbert Braun and Thomas Rieger are research fellows of the Graphical Interactive Systems Group of the Computer Science Department at TU Darmstadt. Thomas Rieger investigates in Virtual Character and Avatar animations with a special view on facial expression and gesture generation by the use of internal mood representation of a Virtual Character. More information on both authors can be found at http://www.gris.informatik.tu-darmstadt.de.
References
[1] Norbert Braun. Automated narration - the path to interactive storytelling. Fraunhofer Publica-
Datenbank, 2002. Workshop on Narrative and Interactive Learning Environments, Edinburgh,
Scotland. [2] Norbert Braun. Narrative semiotics for computer games. Workshop Challenge of Computer Games, Lodz, Poland, 2002.
[3] Vladimir Propp. Morphology of the folktale. International Journal of American Linguistics, 24(4), 1958.
|
Sitemap 