Perception Based Character Modeling and Animation

Ergun Akleman	Mitsutoshi Higa	Tim McLaughlin	Lou Tassinary
Visualization Laboratory Texas A&M University		Industrial Light&Magic	Psychophysiology Laboratory Texas A&M University

Abstract

This sketch summarizes a recent experiment [1] designed to identify a few general rules for altering the perceived sex, gender and attractiveness of humanoid figures. The study is based on prior research in the fields of evolutionary and perceptual psychology. The results of this study can be used for both the modeling and animation of gender specific characters. The results are also informative with respect to the development of user friendly modeling and animation software.

Purpose

Successful animations are typically developed using simple rules for the modeling and animating of expressive characteristics. Examples of such rules are Preston Blair's [2] descriptions of cute, screwball, goofy and heavy characters. These rules, although not evaluated specifically by empirical studies, have proven extremely useful to animators. We believe there is a need for more formal and systematic investigations in order to understand more fully when and why such rules are either effective or ineffective. In this sketch, we investigate a few modeling and animation rules that influence the perception of sex (i.e., Male or Female), gender (i.e., Masculine or Feminine) and attractiveness.

Methodology

Figures with WTH ranging from .9 to .5.

We created twenty-five short animations of a set of walking human silhouettes, varying systematically with respect to body shape and gait, and asked over 350 male and female undergraduates to rate the figures as to their perceived sex, gender and attractiveness.
Different body shapes are obtained by changing the waist-to-hip ratio (WHR) [3] of an otherwise androgynous normal weight human body. We created five WHRs ranging from .5 to .9 in steps of .1, and the actual dimensions of the waist and hip were constrained by data obtained from the U.S. Army Anthropometric survey. Five different gaits were created by varying the relative twist of shoulders and hips of each figure. In general, the shoulders of human males are wider than their hips while the shoulders of females are approximately as same width as their hips; therefore, twisting should be greater at the hips and less at the shoulders for females and greater at the shoulders and less at the hips for males. Put differently, a prototypical female walk is characterized by proportionally greater hip than shoulder movement and an prototypical male walk is characterized by proportionally greater shoulder than hip movement. Different walk cycles are obtained by changing the twist ratio (TR) [4], which was computed by dividing the maximum twist in shoulder motion to the maximum twist in hip motion. Five different gaits were created by using TRs of 0.2, 0.25, 1, 4 and 5, resulting in five distinct gaits that ranged from an exaggerated "swagger" to an exaggerated "sway".

Experimental Results

Walk cycles from exaggerated "swagger" to "sway"

The results of this experiment suggest that the WHR is primarily a cue to the perceived sex of a figure and that the TR is primarily a cue to the perceived gender of a figure. Waist-to-hip ratios of .8 or .9 were nearly universally perceived to be male, whereas WHRs of .5 and .6 were nearly universally perceived to be female. A WHR of .7 was perceived by roughly half the viewers to be male. Twist ratios greater than or equal to 1.0 were perceived as both more masculine and less feminine, whereas TRs less than 1.0 were perceived to be both more feminine and less masculine. Overall, the rated attractiveness of these figures was negative correlated with their WHR. Figures perceived to be male, however, were judged particularly less attractive if their movements were governed by a TR less than 1.0. These relationships provide the basis for rules that might be used effectively by animators to control the perceived sex, gender and attractiveness of humanoid characters. The results of this research might be applied not only to humanoids but also to objects, such as bottles, that have a roughly humanoid form.

Bottles with varying WHR.

Conclusion

The utility of empirically based character modeling and animation is demonstrated based on prior research and the results of a study recently completely in our laboratory. It is argued that the results of this research can and should be integrated into modeling and animation software to afford animators the ability to more directly manipulate parameters designed to effect the perceived sex, gender and attractiveness of characters. The use of similar methods may benefit animators generally by providing more intuitive controls for generating expressive figures and movements.

References

[1] Tassinary, L.G., Higa, M. & Akleman, E. (under review). Cues to sex, gender and attractiveness: The Waist-to-hip ratio and the Twist Index.

[2] Blair, P. (1995). Cartoon Animation, Walter Foster Publishing, Inc.

[3] Tassinary, L.G. & Hansen, K. (1998). A critical test of the waist-to-hip ratio hypothesis of female physical attractiveness. Psychological Science, 9, 150-155.

[4] Tim McLaughlin (1994). The expression of gender in synthetic actors: Modeling and motion control over invariant perceptual cues leading to gender recognition. Unpublished masters thesis, Texas A&M University.