The Birl is an electronic wind instrument which encases a custom circuit with a set of finger pressure sensors, along with a breath pressure and embouchure sensor on the mouthpiece. The device is capable of learning custom sensor-to-synthesis mapping using a neural network and custom training application written in openFrameworks. The synthesis parameters are that of a physical model of a wind instrument, and the audio is generated entirely on a microcontroller in the instrument, or can be sent via OSC to another software application. The hardware was designed by Jeff Snyder.
The software can be found here.
Click on the below image for a closeup.
Training interface and Pedro Eustache playing a trained Birl: