Measuring the Effectiveness of Emotional Digital Humans

Recently, augmented reality has been adopted in industrial environments to address communication inefficiencies and safety risks, yet text- and icon-based AR still lacks human-like interaction. Consequently, integrating AR with digital humans has become increasingly important for enhancing user experience, and advances in artificial intelligence have accelerated efforts to enable more natural and realistic emotional expression. However, conventional digital humans continue to struggle with recognizing and expressing emotions that align with conversational flow orcontext. To address this limitation, this paper proposes a design method and evaluation approach for an LMM-based emotional-support digital human capable of accurately recognizing user emotions and generating natural emotional expressions. The results indicate that such digital humans can express emotions in real time, improving interaction quality and enhancing the overall user experience.