DSPRelated.com
Forums

Interpolation

Started by UD.RaggedRobin November 17, 2009
I seek a program that will take two sound files (A and B) of
approximately equal duration and produce a third sound that is a blend
of A and B throughout its duration. i.e. I want to be able to produce
the middle sound shown below.

AA AA AA AA AA
ab ab ab ab ab <-
BB BB BB BB BB

I'm not looking for what a lot of audio "morphing" programs seem to
do, which is:

AA AA AA AA AA
AA Ab ab aB BB
BB BB BB BB BB

With that said I'm not sure that "blend" is the correct term either -
I don't want the 3rd sound to sound like A and B are simply played
side-by-side. "Interpolation" may be the right term, because the
samples on Ivan Perry's Interpolator page (http://xrl.us/bf7e33 Link
to recherche.ircam.fr) seem to show exactly the process I'm looking
for. Interpolator has another feature I require, which is the ability
to specify a ratio of relative contribution from the original two
sounds. Can anyone suggest how I could obtain a copy of Interpolator
(or perhaps how to contact Ivan), or another program that will do the
same thing? Thank you.
>With that said I'm not sure that "blend" is the correct term either - >I don't want the 3rd sound to sound like A and B are simply played >side-by-side.
So if I'm reading the page correctly, you want the amplitude envelope of one signal and the frequency content of the other?
On Nov 18, 9:01&#4294967295;am, "Michael Plante" <michael.pla...@gmail.com> wrote:
> > So if I'm reading the page correctly, you want the amplitude envelope of > one signal and the frequency content of the other?
Thanks for your response Michael. I'm afraid that I can't answer, as I have no background in signal analysis or audio engineering. (I have a lay understanding of amplitude and frequency, but I don't know what an amplitude envelope is.) I'm interested in a final product that sounds like a smooth blend of the original sounds throughout the duration, as distinct from 2 layered audio tracks such as might be done with a program like Audacity. If the process you describe will produce that, then yes, I'm interested! By way of explanation, I'm looking for a means of presenting blended sounds as experimental stimuli in a discrete trials fading procedure to teach persons with developmental disabilities. This has been done successfully with "morphed" visual stimuli produced in a manner analogous to what I described earlier. We're looking to apply the idea to auditory stimuli.
"UD.RaggedRobin" <ud.ragged.robin@gmail.com> writes:
> [...] > By way of explanation, I'm looking for a means of presenting blended > sounds as experimental stimuli in a discrete trials fading procedure > to teach persons with developmental disabilities. This has been done > successfully with "morphed" visual stimuli produced in a manner > analogous to what I described earlier. We're looking to apply the idea > to auditory stimuli.
Do you have any citations on previous studies in this area? Just curious. Quite awhile ago (20 years?) I saw a special that used a type of random modulation of music to "reach" autistic children. It was one of the most fascinating things I'd ever seen. The video (I believe it was a PBS presentation) I saw used Gordon Lightfoot's "The Wreck of the Edmond Fitzgerald" modulated in amplitude and tone, I believe. -- Randy Yates % "So now it's getting late, Digital Signal Labs % and those who hesitate mailto://yates@ieee.org % got no one..." http://www.digitalsignallabs.com % 'Waterfall', *Face The Music*, ELO
>On Nov 18, 9:01=A0am, "Michael Plante" <michael.pla...@gmail.com> wrote: >> >> So if I'm reading the page correctly, you want the amplitude envelope
of
>> one signal and the frequency content of the other? > >Thanks for your response Michael. I'm afraid that I can't answer, as I >have no background in signal analysis or audio engineering. (I have a >lay understanding of amplitude and frequency, but I don't know what an >amplitude envelope is.) I'm interested in a final product that sounds >like a smooth blend of the original sounds throughout the duration, as >distinct from 2 layered audio tracks such as might be done with a >program like Audacity. If the process you describe will produce that, >then yes, I'm interested!
I don't have a canned solution. Is that what you're looking for? I'm really trying to clarify what it is you want, because your initial explanation sounded exactly like you wanted, to use your words, "A and B [...] simply played side-by-side," at least until I skimmed the link you posted. Unfortunately, your response did not really clarify things much. It sounds like you might need to do some more reading. To a certain extent, there appear to be key terms you can google, including "Super VP", which turned up some interesting graphics on the same page (this URL will wrap; sorry): http://recherche.ircam.fr/equipes/analyse-synthese/DOCUMENTATIONS/perry/spectralenv/spectralenv.html From what I can tell, this abbreviation refers to the term "Super Phase Vocoder" (yes, the letters are apparently transposed), but that phrase only occurs a few times on Google Scholar, and the only article I was able to turn up freely is this one: http://eprints.whiterose.ac.uk/1429/1/cooper.d1.pdf It's not clear that's exactly what you want, but it sounds like it's a step in that direction at least, and they claim the source code (using C on Linux) is available.
I think it would help to do a little *simple* math so your objectives 
will be clearer.  Also, to try to define your terms or to settle on 
terms that you might define for your purposes.

For example, calling the sounds S1 and S2:

Two sounds "side by side" will be heard by a listener in its simplest 
form as:
S1 + S2
this assumes the attenuation from each is equal (and is ignored here) 
and there is no multipath or other strange effects.

Two sound records simply interpolated sample by sample will result in:
(S1 +S2)/2
which is a simple scaled version of the side-by-side case above.  That's 
not much of a difference.

A more sophisticated version of this would be:
(K1*S1 + K2*S2) where K1+K2=1
This would be a "mixer" or a weighted interpolator and from your 
comments you want K1 and K2 to be constant through the segment.

In the Guitar examples it seems like the amplitude is governed by the 
decay of the guitar string vibrations and the sound is governed by the 
other instrument.  That would be a modulation like this:
|S1|*S2
... the magnitude of S1 multiplying S2
so that the sound is that of S2 and the amplitude is that of S1.

and so on .....  I still don't know what you want.

Fred
Fred Marshall wrote:
> I think it would help to do a little *simple* math so your objectives > will be clearer. Also, to try to define your terms or to settle on > terms that you might define for your purposes. > > For example, calling the sounds S1 and S2: > > Two sounds "side by side" will be heard by a listener in its simplest > form as: > S1 + S2 > this assumes the attenuation from each is equal (and is ignored here) > and there is no multipath or other strange effects. > > Two sound records simply interpolated sample by sample will result in: > (S1 +S2)/2 > which is a simple scaled version of the side-by-side case above. That's > not much of a difference. > > A more sophisticated version of this would be: > (K1*S1 + K2*S2) where K1+K2=1 > This would be a "mixer" or a weighted interpolator and from your > comments you want K1 and K2 to be constant through the segment. > > In the Guitar examples it seems like the amplitude is governed by the > decay of the guitar string vibrations and the sound is governed by the > other instrument. That would be a modulation like this: > |S1|*S2 > ... the magnitude of S1 multiplying S2 > so that the sound is that of S2 and the amplitude is that of S1. > > and so on ..... I still don't know what you want.
Apropos nothing much, I am reminded of my brother-in-law when he was 5 years old (and I was not yet in high school). He climbed via a chair onto the dining table and stood there crying inconsolably. My mother-in-law-to-be asked in frustration, "What do you /want/? He answered, "Something else!" Jerry -- Engineering is the art of making what you want from things you can get. &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;