Music Information Robotics: Coping Strategies for Musically Challenged Robots.

In the past few years there has been a growing interest in music robotics. Robotic instruments that generate sound acoustically using actuators have been increasingly developed and used in performances and compositions over the past 10 years. Although such devices can be very sophisticated mechanically, in most cases they are passive devices that directly respond to control messages from a computer. In the few cases where more sophisticated control and feedback is employed it is in the form of simple mappings with little musical understanding. Several techniques for extracting musical information have been proposed in the field of music information retrieval. In most cases the focus has been the batch processing of large audio collections rather than real time performance understanding. In this paper we describe how such techniques can be adapted to deal with some of the practical problems we have experienced in our own work with music robotics. Of particular importance is the idea of self-awareness or proprioception in which the robot(s) adapt their behavior based on understanding the connection between their actions and sound generation through listening. More specifically we describe techniques for solving the following problems: 1) controller mapping 2) velocity calibration, and 3) gesture recognition.