Perception of Micro-Expressions in Animated Characters with Different Visual Styles

Tianyu Hou, Purdue University

Abstract

The purpose of this research was to examine the perception of micro-expressions in animated characters with different visual styles. Specifically, the work reported in this thesis sought to examine: (1) whether people can recognize micro-expressions in animated characters, (2) whether there are differences in recognition based on the character visual style (stylized versus realistic), (3) the extent to which the degree of exaggeration of micro-expressions affect the perceived naturalness and intensity of the animated characters’ emotion, and (4) whether there are differences in effects on perceived naturalness and intensity based on the character visual style. The research work involved two experiments: a recognition study and an emotion rating study. A total of 275 participants participated in both experiments. In the recognition study, the participants watched eight micro-expression animations representing four different emotions (happy, sad, fear, surprised). Four animations featured a stylized character and four a realistic character. For each animation, subjects were asked to identify the character’s emotion conveyed by the microexpression. Results showed that all four emotions for both characters were recognized with an acceptable degree of accuracy. The recognition rates of the stylized character were 84.73% for happiness, 88.73% for sadness, 60.73% for fear, and 83.64% surprise. The recognition rates of the realistic character were 87.37% for happiness, 82.94% for sadness, 69.62% for fear, and 77.13% for surprise. In the emotion rating study, participants watched two sets of eight animation clips (16 clips in total). Eight animations in each set featured the character performing both macro- and micro-expressions, the different between these two sets was the exaggeration degree of microexpressions (normal vs exaggerated). Participants were asked to recognize the character’s true emotion (conveyed by the micro-expressions) and rate the naturalness and intensity of the character’s emotion in each clip using a 5-point Likert scale. Findings showed that the degree of exaggeration of the micro-expressions had a significant effect on emotion’s naturalness rating, emotion’s intensity rating, and true emotion recognition and the character visual style had a significant effect on emotion’s intensity rating. Emotion type, participant gender and participant animation experience also had significant effects on perception of the micro-expression.

Degree

M.Sc.

Subject Area

Design

COinS