Hi,
I'm trying to find a way to analyze human speech, in terms of its formants. My intention is to create an accompanying sound that will follow the speech, based on the analysis parameters. Honestly, I know very little about human speech, and I'm going to look into this matter. But, meanwhile, [fiddle~] seems to be very limited to this kind of analysis. Any suggestions or referrals where to begin?
thanks,
yMotion

check out [fft~] and [rfft~]. those are the two main tools for spectral analysis, afaik. I also recommend arss.

Hi Sunji
Thanks a lot. The arss looks very interesting. I'm looking into it now.

I wonder if you've seen this external library.

http://williambrent.conflations.com/pages/research.html

I was thinking of timbre ID. It has a nice speech recognition object. I used it (on Mac OSX Lion - no issues) on a project where I sent speech simultaneously to [fiddle~] and the speech recognition and used them simultaneously to control the parameters of a synth. Which sounds fairly close to the type of project you're doing.

These are the ready to hand solutions I can think of. But if you want to build it yourself I believe you could begin by looking at the vocoder and timbre stamp patches in the PD help browser (I06. and I07.)

Also, if you need to make one audio signal sound like another (not sure if this is what you mean or importantly if it should be real-time) then [convolve] from the same library I mentioned or the timbre based granular synth could work well.

Hi Jermain,
Thanks a lot! I'll look into this library. And yes, your project sounds like something similar to mine. If you have any documentation or sound-files you want to share I'll be happy to see/listen to.
ymotion