Toward Generative Sound Cues for Robots Using Emotive Musification

Abstract

Sound is an essential yet under-studied way for robots to communicate with humans. When designed well, robot sound can improve aspects of human-robot interaction from social perception to team fluency, but at its worst, robot sound can discourage use of robots altogether. Thus, sound cues must be carefully and intentionally designed. To address this need, we present a system that uses the inherent emotional connotations of musical qualities to dynamically generate emotive sound for robots. Our application utilizes real-time modification of tempo, pitch, scale, and sound brightness to algorithmically generate melodic phrases intended to evoke specific moods or feelings in human listeners. An in-the-wild exploratory study with N = 26 participants demonstrated that our generative sounds caused human listeners to perceive the robot as happier and warmer. This effort is a first step toward a planned full system that will democratize the design of music-based emotional communication in human-robot interaction.