A tension-moderating mechanism for promoting speech-based human-robot interaction

Abstract

We propose a method for promoting human-robot interaction based on emotion recognition with particular focus on tension emotion. There are two types of emotions expressed in a short time. One is autonomic emotion caused by a stimulus, such as joy and fear. The other is self-reported emotion, such as tension, that is relatively independent of a single stimulus. In our preliminary experiment, we observed that tension emotion (self-reported emotion) obstructs the expression of autonomic emotion, which has demerits on speech recognition and interaction. Our method is based on detection and moderation of tension emotion. If a robot detects tension emotion, it tries to ease it so that a person will interact with it more comfortably and express autonomic emotions. It also retrieves nuances from expressed emotions for supplementing insufficient speech recognition, which will also promote interaction.