Over the last year, I have concentrated on new solo work for piano and/or computer.
I am developing methods to control the computer through acoustic information, and through game controllers with haptic feedback.

The acoustic input is categorized in bins based on similarity discovered with self-learning neural networks.The bins are processed with various effects. I am searching for pleasant unpredictability, a kind of serendipity in fact.
The problem is to find useful feature-vectors that takes little computational overhead to calculate. This is made a little easier by the fact that I am not looking to emulate human understanding of sound.

The computer sounds are completely synthesized. All rhythms are played/generated in the moment.They are structured with an abstracted pattern-sequencer that can generate recursive patterns where each step can trigger time-scaled versions of the parent pattern. It is controlled with a force-feedback joystick, so that each type of sound has a different "feel" in the joystick. As an added bonus the joystick will move by itself.