|
(UK) Dan Stowell (MCLD) is a musician, programmer, and scientist. He is currently studying for a PhD at Queen Mary University of London, where he is developing techniques to use voice sounds such as beatboxing to control electronic instruments. Musically he combines live-coding with beatboxing - solo and as part of the duo Spoonfight - and also produces installation work as part of C4DM Presents. He's also one of the current lead developers of SC. http://www.mcld.co.uk/
Timbre Remapping with Regression Trees
When using the timbre of a signal to control some other system (e.g. using voice to control a synthesiser), the question arises of how a machine could learn to create timbral "analogies" - complicated by timbre's multidimensional nature. We present our regression tree technique which learns associations between two unlabelled multidimensional distributions, and apply the technique to a simple timbral concatenative synthesis system.
SuperCollider and Android
Dan Stowell / Alex Shaw
SuperCollider's audio server has great potential as an audio engine for mobile applications, and Android offers an interesting mobile OS for a variety of devices. Dan and Alex will describe the Android infrastructure and how SuperCollider's architecture fits into this ecosystem. Then they will describe their port of scsynth to Android, demonstrating some of the applications they have created, and how you can do interesting sound things on Android right now.
When using the timbre of a signal to control some other system (e.g.
using voice to control a synthesiser), the question arises of how a
machine could learn to create timbral "analogies" - complicated by
timbre's multidimensional nature. We present our regression tree
technique which learns associations between two unlabelled
multidimensional distributions, and apply the technique to a simple
timbral concatenative synthesis system.When using the timbre of a signal to control some other system (e.g.using voice to control a synthesiser), the question arises of how amachine could learn to create timbral "analogies" - complicated bytimbre's multidimensional nature. We present our regression treetechnique which learns associations between two unlabelledmultidimensional distributions, and apply the technique to a simpletimbral concatenative synthesis system.
|
