Jamie-
I ended up using pyext in the end, which worked wonderfully. I would write each fft signal block into an array, then use the numarray stuff in pyext to iterate through each block, calculate the centroid and output a float. My vanilla PD implementation never seemed to work quite properly, and I started playing with pyext before I even figured it out... I will post all this stuff shortly when I have a second to sort it out.
When I have some time, I'd like to plug some of the other formulae you provided into my setup and see what can happen. Using a technique like reinforcement learning for musical purposes might have some promise, but it seems like the side effects are more interesting than the intended ones...which puts it on strange ground, but isn't that what we are doing anyway?
Jacob
On 10/17/05, Jamie Bullock jamie@postlude.co.uk wrote:
Just out of interest, how did you end up calculating the centroid?
Jamie
On Sun, 2005-10-16 at 21:18 +0200, Jacob Last wrote:
I've also got a set of granular synth patches from my work this past summer, which I was working on integrating with a reinforcement learning system that used the spectral centroid of the output to control the synth parameters. I will post it in a few days once I have a chance to make sure it works independently and cleanly.
Best Jacob
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list