[music-dsp] matching two waveforms for subtraction

[Aaron Brick]

hello svante,

thanks for writing. is cross-correlation the only, or just the preferred,
method for doing this? and do you have any reference to an implementation?
all i can find are complex equations; my intention is to use the technique
in GNU octave (akin to matlab).

as long as the BPM is constant, even with some error introduced by human
playing, i imagine that results would be passable. as far as the high
frequencies you mention, since they will never be present in speech, i can
just roll response off after 5 or 10 khz to eliminate hihats and such,
correct?

thanks much,

aaron.

so said svante_dsp at hotmail.com in 1.9K bytes at Tue, Aug 07, 2001:

> 
> I've tried this myself, with varying results. Cross correlation gives some 
> good results if the original and the instrumental are identical apart from 
> the presence of vocals. If a track for example is done with pro tools, so 
> that there is no pitch jitter or midi ditto, then this should work fine. 
> Tracks done on analog tape is a whole different story.
> 
> Another problem is that frequencies near Nyquist are hard to get at, so if 
> f=44100, then you'll probably won't get rid of hihat sizzles and such.
> 
> /Svante
> 
> >From: Aaron Brick Reply-To: music-dsp at shoko.calarts.edu To: 
> >music-dsp at shoko.calarts.edu Subject: [music-dsp] matching two waveforms for 
> >subtraction Date: Tue, 7 Aug 2001 01:48:56 -0700
> >
> >hello all,
> >
> >i wonder if you all have any input as to techniques i could use to 
> >accomplish this rather simple DSP operation. it occurs to me that i would 
> >like to take samples of some sung lyrics, but without the accompanying 
> >music. since i can easily sample a stretch of near-identical music without 
> >the vocal, it seems that by subtracting it from the other wave i could come 
> >close to isolating the voice. however, this obviously requires lining up 
> >the two samples correctly. one's ear can do it quite well, of course.
> >
> >does anyone know of a way to detect certain trends in two waveforms as i 
> >decribe?
> >
> >thanks,
> >
> >aaron.
> >
> >- | | | | | | | | | | - - - - - - aaron brick - - aa at lithic.org - - - - - - 
> >| | | | | | | | | | -
> >
> >
> >dupswapdrop -- the music-dsp mailing list and website: subscription info, 
> >FAQ, source code archive, list archive, book reviews, dsp links 
> >http://shoko.calarts.edu/musicdsp/
> >
> 
> _________________________________________________________________
> Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp
> 
> 
> dupswapdrop -- the music-dsp mailing list and website: subscription info,
> FAQ, source code archive, list archive, book reviews, dsp links
> http://shoko.calarts.edu/musicdsp/
> 


                           - | | | | | | | | | | -
                           -                     -
                           -                     -
                           -     aaron brick     -
                           -    aa at lithic.org    -
                           -                     -
                           -                     -
                           - | | | | | | | | | | -


dupswapdrop -- the music-dsp mailing list and website: subscription info,
FAQ, source code archive, list archive, book reviews, dsp links
http://shoko.calarts.edu/musicdsp/