Re: Fundamental Pitch Filter

BlueXIV can speak for himself about his specfic requirements but ";pitch"; has a specific well defined meaning  in music which I expect most people on this list know. It means the frequency which defines the note an instrument(in this case the voice) appears to be playing. Pitch is usually expressed in terms of notes in a scale but frequency is also valid, especially if you are looking for differences or errors. I say ";appears to be playing" because there can be some subjectivity involved which you might as well call audio illusions but most often pitch means the frequency of the lowest harmonic of an instrument sound. People are often sloppy about mixing the terms pitch and frequency. Technically pitch is a perceptual (happens in the ear) phenomenon and frequency is purely a physical property of the sound. The sloppiness comes because they are closely coupled. In a little more detail pitch is usually calculated as the frequency of the lowest harmonic of a sound which has significant energy in it. "Significant energy" is a subjective term but it mostly means the harmonic is loud enough not to be drowned out by the higher frequency harmonics. The voice as an instrument has a complex harmonic structure with lots of variations and it is harder to specify a reliable calculation of the pitch of a sung note.

This should not be surprising because the voice can credibly imitate many instruments. Making one algorithm that covers all of them wouldn';t be easy. For example I am learning to sing vocal percussion and one of the big surprises to me is that even when I make a sound that "sounds like" a particular percussion instrument the spectra of the instrument sound and the spectra of my imitation are quite different. The surprise to me is that "sounds like" does not mean "has similar spectra" which had been my assumption. The apparent reason is that what we hear is the result of both the physics of the sound detectors in the ear plus the results of some post processing that gets done on our ear's detector output signals before it reaches the perceptual levels of the brain. Just like there are an infinite variety of optical spectra that result in the same color there must be a wide variety of sounds that sound "similar." Mathematical pitch detection really needs to model the ears sound perception system. Unfortunately that is not completely understood yet.

The real lesson is to watch out for apparently simple words like pitch. The meaning is clear in an orchestra or chorus because we use our ears to hear it. That works because pitch is defined in terms of what we hear. It is a lot less clear for tools like Audacity because the bio-mechanics and psycho-acoustics of pitch detection are not completely understood so pitch is not mathematically well defined. We are usually dealing with mathematical approximations (like my definition above) that work only with a subset of instruments.

Here is an extreme but very real and interesting example that probably happens in your very own computer. It is possible to make people hear a low fundamental harmonic that doesn't exist by filling in the upper harmonics usually associated with it. This is done on modern sound cards with dsp processors. They analyze the sound to find base frequencies to boost. Instead of physically boosting these frequencies (smaller computer speakers don't work well at those frequencies) they add in the upper harmonic pattern instead resulting in in apparent bass sound that the small computer speakers can't actually make. This apparently boosted bass sound is an audio illusion which tricks the ears into hearing what is not there resulting in stronger bass than would otherwise be possible. Every sound card seems to have patents and trademarks associated with variations on this trick. I've heard (but am not sure I believe) that vocal quartets particularly good at close harmony use this (probably without realizing the details) to make chords seem to have 5 or more parts.

It might be interesting to create a plugin which does this.

Try to calculate the pitch of that sucker!

"The other day singing in the  choir
I heard a pitch that wasn't there.
It wasn't there again today.
I wish, I wish it 'd go away.";
(Apologies to William Hughes Mearns.)

Anyway by a "series of fundamental pitches" I'm guessing that BlueXIV really means a plot of pitch vs. time rather than a spectrum but as I said he should really say for himself what he means.

Regardless of what BlueXIV wants I would like to know if Audacity can display a plot of pitch vs. time. It would probably only have meaning for a solo recording. (Calculating the pitch of multiple voices is much harder.) I seem to remember seeing such a thing once but I can't find it in the menus. If there is such a thing (assuming I was not smoking something particularly fine that day) does anyone know what kind of algorithm it uses to calculate pitch?

I've been trying to create such a pitch tracking plugin based on the pitch tracking algorithm from the Tartini program (open source) but I don't want to duplicate effort if something similar has already been done. I really want to use it, not write it.

On Wed, Mar 19, 2008 at 8:36 AM, Wes Morrison < wesmorrisonjuno.com" target="_blank">wesmorrisonjuno.com&gt; wrote:

I&#39;m not exactly sure what you mean by "a series of only the Fundamental
Pitches." Do you mean frequency analysis, or processing the sound in some
way? I assume the second because Audacity already has Analyze/Plot Spectrum.


BlueXIV wrote:
&gt;
> Hello everyone,
>  I am a new user to audacity, and I'm really familiar with much of the
> audio editing functions it supports. I was wondering if anyone here knows
&gt; how to create an effect that will transform a sound file (I plan to use a
> short speech .wav) into a series of only the Fundamental Pitches? I know
>; Audacity can find these pitches, but I'm not familiar enough with the
> program to actually make it on my own. If anyone could help, it'd be much
>; appreciated.
>

--
View this message in context: http://www.nabble.com/Fundamental-Pitch-Filter-tp16136432p16143527.html
Sent from the audacity-nyquist mailing list archive at Nabble.com.


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Audacity-nyquist mailing list
Audacity-nyquistlists.sourceforge.net" target="_blank">Audacity-nyquistlists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/audacity-nyquist

--
Steve Morris
barbershopstevegmail.com" target="_blank">barbershopstevegmail.com
Bass: Unnamed quintet/quartet whatever
Bass: Sounds Of Concord