i think what Joe means by this is what is commonly referred to as a vocal remover. meaning it takes out the center channel, and i believe that it's done using M/S processing on the stereo signal. i can tell you it likely won't involve CPU intensive tasks like FFT because they used to sell kits to remove vocals in the 1980s:
http://www.paia.com/ProdArticles/vocalzap.htm
Ahh ok sure. Yeah you're right I think, it's annoying because in theory, L - R should give you what's not in the centre. But that ends up not being very useful anyway :)I'm no expert, but sounds like Mid-Sides could give you what you need, no? then you can get rid of the Mid and voilą : sides.Eran, Mid-side is a different encoding to Stereo as you have 3 channels instead of 2, meaning the mono signal is already available.Cheers,JoeOn 12 October 2012 10:49, i go bananas <hard.off@gmail.com> wrote:oops, sorry about the double reply Joe, i forgot 'reply to all'
to turn a stereo signal into a mono one, yes it's L+R..but that's not what i wanted to do. I wanted to take ONLY the parts of a stereo mix which lie right in the middle. It would be basically the inverse of a stereo difference effect.
To put it another way, Stereo difference removes those parts of the mix which are panned dead center. What i want to do is leave those parts, and then remove anything that isn't panned dead center.
At first, i thought it would be as simple as just subtracting the stereo difference signal from the original mix. But when i actually tried it, it didn't work, and then also when i worked it out logically on paper, i realized that it's impossible just with simple arithmetic.
My best guess, is that you could FFT the stereo difference signal to get it's spectral profile, and then subtract that from the original mix. Might give it a go some time.
--
Follow me on Twitter @diplojocus
_______________________________________________
Pd-list@iem.at mailing list
UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list