4

Is there any free software tool or combination that allows me to identify the pitch of a recorded singing session?

The idea is to display some kind of graph with the current pitch in a time line along with markers for the standard notes (C3, C#3, D, etc). I don't need pitch correction and I don't need it to be done in real time, either.

I know that once there was a plugin for Rosegarden that did that, but it has gone missing.

skynet
  • 9,898
  • 5
  • 43
  • 52
Luis Soeiro
  • 4,262
  • 6
  • 35
  • 45

4 Answers4

5

Checkout Audacity. It came out of a project to do musical pitch analysis.

EBGreen
  • 36,735
  • 12
  • 65
  • 85
  • Listen to http://www.twit.tv/floss42 for some insights into what Audacity can do. I think it can do this, or can be convinced to do it. It sounds like something that could already exist. – jakobengblom2 Oct 01 '08 at 19:56
  • That podcast is what prompted me to suggest it. :) I've never done this with audacity myself, but after listening to that I can't believe that it won't do it. – EBGreen Oct 01 '08 at 20:10
0

Not exactly what you are looking for, but the Singstar lookalike Ultrastar-NG at least does something like this. http://ultrastar-ng.sourceforge.net/

matli
  • 27,922
  • 6
  • 37
  • 37
0

I'm unaware of any software package that has this built in. If you're interested in writing something like this, you'll want to look at Discrete Fourier Transforms. This turns a time-series sample into a collection of frequencies. But this leaves you with no information about when the various frequencies occur, so you must do a windowed Fourier Transform, with windows of whatever time-resolution you want. Increasing the time resolution decreases the frequency resolution, however.

The simplest thing to do is to figure out the largest frequency component in any window and call that the frequency. But real music (a) has chords and (b) has overtones and undertones. In addition singing often has "tremolo", where the singer varies the actual pitch around the theoretical pitch the music is marked at.

wnoise
  • 9,764
  • 37
  • 47
  • I meant to use it with human voice, without other instruments. I know that commercial packages like Antares do it somehow. – Luis Soeiro Oct 01 '08 at 20:51
  • 1
    This isn't as easy as it seems. Look at this undergraduate project (happens to be mine...): http://www-sipl.technion.ac.il/new/Archive/Special_Events/sipl2004/Projects_PowerPoint/WAV-to-MIDI.pdf – Nathan Fellman Nov 24 '08 at 13:43
  • DFT isn't the right answer for pitch analysis, the proper algorithm to use is called Autocorretion (http://en.wikipedia.org/wiki/Autocorrelation) and it's what most of the big packages use. – slacy Apr 30 '15 at 23:20
  • @stacy: Autocorrelation is most efficiently computed using the DFT, unless the window size is extremely small. Autocorrelation also leaves the result in the time domain, but he explicitly wants the frequency domain. – wnoise Jul 02 '15 at 01:14
0

Praat will at least do automatic pitch estimation of complex sounds. Though I don't know if it can mark the standard notes as you requested.

Rob

RobS
  • 3,807
  • 27
  • 34