A Study on the Interaction Between Humanoid Robots Equipped with Korean Dataset-Based Gesture Generation AI Models and Korean Users

Humanoid robots can interact effectively with humans through non-verbal expressions such as gestures. To generate non-verbal expressions such as gestures for humanoids, gesture generation AI models and datasets are essential. However, most publicly available gesture generation datasets are based on English and may not reflect the linguistic and cultural characteristics of non-English-speaking regions like Korea. To address this, we built a Korean-specific dataset and trained a gesture generation AI model. We implemented the humanoid robot AIMY in a simulation, applied the generated gestures to the AIMY, and conducted demonstrations with Korean users to evaluate the impact on human-robot interaction. The results demonstrate that gestures trained on Korean datasets better align with Korean non-verbal expressions and are preferred by Korean users compared to gestures based on English datasets, enhancing user experience and interaction quality.