SQUISHBOI: A Multidimensional Controller for Complex Musical Interactions using Machine Learning

We present SQUISHBOI, a continuous touch controller for interacting with complex musical systems. An elastic rubber membrane forms the playing surface of the instrument, while machine learning is used for dimensionality reduction and gesture recognition. The membrane is stretched over a hollow shell which permits considerable depth excursion, with an array of distance sensors tracking the surface displacement from underneath. The inherent dynamics of the membrane lead to cross-coupling between nearby sensors, however we do not see this as a flaw or limitation. Instead we find this coupling gives structure to the playing techniques and mapping schemes chosen by the user. The instrument is best utilized as a tool for actively designing abstraction and forming a relative control structure within a given system, one which allows for intuitive gestural control beyond what can be accomplished with conventional musical controllers.