Validation of vocal prosody modifications to communicate emotion in robot speech

Abstract

This research investigated the use of Mary, an open source speech synthesizer, to convey the emotional intent of a robot through the robot's vocal prosody. The robot's pitch, pitch range, speech rate, and volume were varied to convey anger, fear, happiness, and sadness. The results indicate participants recognized the intended emotions at a rate statistically higher than chance whether they provided their own free choice word or used a forced choice survey to describe the emotion presented by the robot. As expected, the participants correctly transcribed more of the words spoken by the robot when the vocal prosody modifications performed were small (sadness) than when the vocal prosody modifications were large (anger and fear).