Online processing for speech-driven gesture motion generation in android robots

Hand gestures commonly occur in daily dialogue interactions, and have important functions in communication. In this study, we proposed and implemented an online processing for a speech-driven gesture motion generation in an android robot dialogue system. Issues on motion overlaps and speech interruptions by the dialogue partner were taken into account. We then conducted two experiments to evaluate the effects of occasional dis-synchrony between the generated motions and speech, and the effects of holding duration control after speech interruptions. Evaluation results indicated that gestures should not be delayed by more than 400ms relative to the speech utterances. Evaluation of the second experiment indicated that gesture holding durations around 0.5 to 2 seconds after an interruption look natural, while longer durations may cause impression of displeasure by the robot.