Music to Motion: Using Music Information to Create Expressive Robot Motion

Abstract

In a robot theater, developing high-quality motions for a humanoid robot requires significant time and effort. People with artistic and animation skills are required to create expressive gestures and movements on the robots. We observe that dancers can translate music into motions that are perceived as expressive by the audience. In this study, we developed a program to exploit melodic and dynamics information from music to create expressive robot motion. The program was used on two applications: producing motions strictly from the music information and controlling the execution a pre-programmed sequence of actions. Both applications are intended for a robot theater. The former was applied on an arm robot, while the latter on a humanoid KHR-1 robot. Two surveys were done to analyze the impact of our method. The results suggest that the robot motions produced by our program can be perceived as being more expressive and more dynamic than the motions created by a person or without music information.