Breathe and Speak Attentively: Implementing Respiratory Awareness into Conversational Robots

This study explores the use of respiratory information to enhance human-robot spoken interactions. We implemented two functions: 1) speech collision avoidance (SCA) based on speech onset prediction using human respiratory information and 2) pseudo-respiration presentation (PRP) to humans by the vertical movements of a robot. Experiments involving 26 participants confirmed that SCA and synchronous PRP aligned with human respiration effectively suppressed speech collisions. Additionally, the synchronous PRP enhanced the perceived uninterruptedness of interactions and showed a tendency to increase some impression scores. However, our findings suggest that preventing speech collisions alone does not necessarily improve interaction experiences, highlighting the need for more sophisticated strategies to enhance human-robot spoken interactions. Moreover, our findings indicate that enhancing compatibility between humans and robots is a challenging task and should be actively addressed to further improve human-robot spoken interactions.