Home // International Journal On Advances in Life Sciences, volume 14, numbers 3 and 4, 2022 // View article

A Study on the Quality of Facial Expression in Digital Humans

Authors:
Shiori Kikuchi
Hisayoshi Ito
Oky Dicky Ardiansyah Prima

Keywords: digital human; facial expression; basic emotions; avatar

Abstract:
There have been many examples recently of the use of digital humans for interactive communication, such as in customer support and digital health care. The use of digital humans is expected to reduce communication barriers caused by differences in personal appearance, behavior, and facial expressions. Despite the potential for digital humans to represent realistic nonverbal communication, there is a lack of evidence regarding the extent to how effective they are in communication. This study attempts to evaluate the six basic emotions (anger, fear, disgust, happiness, sadness, and surprise) of digital humans with the following two experiments. First, the quality of facial expression is evaluated by an actor's facial expression and the expression of the digital human created from the actor. Second, we evaluate the quality of the facial expressions of 17 subjects captured from various angles and those of digital humans created from the corresponding angles. Two deep learning-based facial expression recognitions: DeepFace and HSEmotion were used for the evaluation. Experimental results showed that HSEmotion demonstrated a more stable recognition rate than DeepFace for the same facial expressions of subjects captured from different directions. However, when the facial expressions of these subjects were transferred to the digital humans, both tools failed to properly recognize their facial expressions. Future work will include a facial expression recognition library that considers both real people and digital humans.

Pages: 110 to 119

Copyright: Copyright (c) to authors, 2022. Used with permission.

Publication date: December 31, 2022

Published in: journal

ISSN: 1942-2660