Stompboxes: Kicking the Habit

Burlet, Gregory and Fujinaga, Ichiro

Proceedings of the International Conference on New Interfaces for Musical Expression

Sensor-based gesture recognition is investigated as a possible solution to theproblem of managing an overwhelming number of audio effects in live guitarperformances. A realtime gesture recognition system, which automaticallytoggles digital audio effects according to gestural information captured by anaccelerometer attached to the body of a guitar, is presented. To supplement theseveral predefined gestures provided by the recognition system, personalizedgestures may be trained by the user. Upon successful recognition of a gesture,the corresponding audio effects are applied to the guitar signal and visualfeedback is provided to the user. An evaluation of the system yielded 86%accuracy for user-independent recognition and 99% accuracy for user-dependentrecognition, on average.