An Empirical Study on Emotion Recognition and User Consensus of Full-Body Gestures in Metaverse Environments

Social virtual reality (VR) platforms increasingly rely on avatars as primary interfaces for social interactions, with body movements serving as key channels for emotional expression. While previous studies have examined emotion recognition, the gestures users consider most representative remain underexplored. This study investigated recognition accuracy and user agreement for 30 full-body gesture variants representing six basic emotions. In a large-scale perceptual evaluation (N = 187), participants completed recognition and representative gesture selection tasks. All gesture variants were recognized at rates above chance level, demonstrating that emotional gestures derived from prior research retain recognizable meanings in VR avatar contexts. Recognition accuracy and user agreement did not always correspond, and LMA-based confusion analysis revealed shared motion characteristics among frequently confused emotions. An example application illustrates the implementation of well-recognized gestures in real-time avatar-mediated interactions. These findings provide practical guidance for selecting avatar emotional gestures aligned with user expectations in social VR systems.