Towards an FMG based augmented musical instrument interface

Studies suggest that introducing musical activities in physical rehabilitation protocols has the potential to enhance the outcome for users who suffered from neurological disorders or direct physical injuries. One kind of activities is to allow users to exercise using musical instruments. However, due to their physical limitations, this option may not be feasible. In order to provide the users with such an option, we propose a novel gesture control interface which allows a user to play a virtual musical instrument based on an emerging technique called force myography (FMG). FMG detects muscle activity pattern based on multiple force sensors surrounding a limb. Using machine learning algorithms, the registered FMG pattern can be associated with different commands to control external devices. Based on this idea, we developed a virtual piano which can be controlled by a wrist strap with FMG and motion sensors. We evaluated the current system based on the offline classification performance of FMG to detect three gestures plus the default relaxed state. The three gestures were associated with producing a single note sound, a major chord sound, and a minor chord sound. A preliminary single evaluation trial showed these gestures can be predicted with 90% accuracy. Using these gestures, we conducted an online testing in which a user played a musical tune with 41 keynotes on the virtual piano. The user was able to complete the tune within 3 minutes with 3 false predictions. The result demonstrated the feasibility of using FMG to control a virtual musical instrument; however, more research and development are needed to improve the performance and usability of the system.