Theory behind Pitch Correction Software

[DevonTone]

Hi!

I'm sitting on my bachelor thesis right now in which I should try to find out what's happening behind the scenes of pitch correction programs.

Actual Title: Comparison of current proprietary pitch correction software with respective freeware solutions and a technical forecast on future possiblities in sound manipulation

I want to try and find out how these programs work by using analysis software... if that's even possible.
To be honest, I still have no real idea how to get this working.

For quite a while I'm looking now for literature that could help me.
I found a whole bunch of "the perfect vocals tracking" or similar "recording tutorials", but nothing that specifically addresses this issue and would give me a rough guide.

Focus of the work will be clearly Antares Auto-Tune and Celemony Melodyne. Likewise, I want to test how open-source programs are good in this respect.
With Melodyne I had a lot to work with by now and would consider myself as quite competent. Auto-Tune on the other hand is completely new territory.

The forecast would pretty much be my personal prediction and just what can be found in any literature about it.
In addition to that I'd like to go into the multitimbral aspects of melodyne and probably Prosoniq sonicWORX.

In my web search I just found articles that deal with older versions and were pretty much limited to the performance of these softwares.
The archive search on websites of relevant journals hasn't been very useful either.

I would be really grateful if somebody could give me any advise how to deal with this problem. Some literature advices, some hints for web articles, some "...don't get started on this, won't be much of use..."

Thank you in advance!

[The Byre]

In its' simplest form - imagine a tuner linked to an analysis and operating routine, linked in turn to a harmoniser. Tuner registers x deviation, tells the routine and it tells the harmoniser to raise or lower the pitch in the usual way as little samples overlapping one another and played back slightly faster or slower.

But that ain't how Melodyne works. There was a paper on it somewhere, I'll have a dig . . .

[DevonTone]

But that ain't how Melodyne works. There was a paper on it somewhere, I'll have a dig . . .

Thanks for your fast reply!
If there's a paper about Melodyne I'd love to get a hold of it.

Well...
Just noticed that this might be the wrong sub forum for this thread.
Sorry Mods... I'm new here
Is there a way to move it to a part that concerns these tye of questions more likely?
Thanks!

[Dave C]

Your title is too long. Way too long.

What is wrong with "comparison of pitch correction software"?

It is easier to write, to read, and you can then take it whatever way your research leads you.

Is this a bachelor of science or bachelor of arts degree you are pursuing?
Your original title is more appropriate for a BA degree. For a BSc, you would be expected to investigate and discuss the technical methods used, and to show a full understanding of them. This very likely could not be deduced from using analysis software (what kind of software are you thinking of using?) or would be very difficult to work out.

Many pitch corectors work by taking a signal which is a varying voltage in the time domain, or a digital representation of such, aplying an FFT to get it into the frequency domain, and then shifting the FFT spectrum up or down by the appropriate amount, before then applying an inverse FFT to shift the frequency spectrum to get it back into the time domain as a digital representation of a time varying voltage.

You could also look at 'frequency warping' which is a non linear shift of the spectrum.

appropriate research terms would be 'fast fourier transform' and 'phase vocoder'.

Read 'the computer music tutorial' by Curtis Roads.



@ Byre - I'd like to read the paper on melodyne if you can find the link. Many Thanks.

@ Original poster - Good luck with your studies,

Dave.

[DevonTone]

Hi Dave,

thanks!
I'll have a look at Curtis Roads.

It's a Bachelor of Engineering in Mediatechnoligies. Probably should have mentioned that. Sorry.
The title was given to my by my professor. Would have loved to put it in a more simple form.

Not sure about the software I want to use yet.
Found these in the web:

Raven: Interactive Sound Analysis Software
Sigview spectrum analyzer - FFT based signal analysis software

Both cost money but offer a full-working trial for 30 days.
Maybe this limitation puts some healthy pressure on me

In my signal processing course we were working with dasy lab:
www.dasylab.com
Should have that program somewhere around.

By now I'm searching university archives. Maybe somebody already wrote something about this topic.