Abstract
The global outbreak of COVID-19 that started in 2019 made face-to-face communication challenging. As one alternative, attention was shifted to the metaverse, virtual reality (VR) space on the Internet. Particularly, avatar communication using avatars garnered attention. Avatar communication aids diverse nonverbal information to be given to avatars and is presumed to be applied in different fields including education. Nevertheless, many existing metaverse systems have considerable environmental constraints to function. Additionally, the complex environment required to include nonverbal information makes it challenging for many people to use. The final objective of this study is to construct a metaverse system that can easily add nonverbal information. As an intermediate step, we executed the extraction of open/closed eyes by applying MediaPipe. The outcomes demonstrated a 86.70% increment in the percentage of correct responses in the case of wearing glasses compared to the prior investigation. Additionally, the avatar was able to move following the extracted data. Still, because of the small number of experimental situations, it is necessary to confirm whether the system is general-purpose or not. It is also necessary to integrate the system into a metaverse system to validate the actual operation.
