In an era of increasingly immersive digital environments, human-avatar interaction must evolve to accommodate the full spectrum of human diversity. This project proposes a novel multimodal AI framework that leverages voice, facial expressions, gestures, and contextual cues to create emotionally intelligent and accessible avatars. By integrating advanced deep-learning techniques with real-time perceptual feedback, the system adapts to diverse user needs—including those with visible and invisible disabilities—ensuring inclusive, empathetic, and natural interaction. Grounded in a multidisciplinary review of current advances in virtual embodiment, non-verbal communication, and accessible AI design, our approach aims to redefine avatar systems as not only functional but also socially and ethically responsive. The outcome will contribute to the development of inclusive digital ecosystems where every individual can interact, express, and engage with authenticity and dignity.
Keerthana et al. (Mon,) studied this question.