by the way have you seen this project:  http://www.youtube.com/user/NVRS242#p/u/4/prIOW1gKFtw ?  

Extracting pitch from a melody (humming, whistling, as they do on that video) is a bit easier than what I'm trying to do, which is extracting pitch from a human voice. That is, from someone who is speaking, not singing! When you sing, you usually maintain pitch for longer times than when you speak. Plus, knowing the target notes (A, A#, B, C, C#...) introduces a priori knowledge (you know the exact frequency for each note). I don't know if I could use something similar when studying human voice :(

The thing is that I'm finding Processing not exactly suitable for my needs. I'm not sure if it's fast enough to handle pitch detection and all the other things that my project requires (drawing on screen certain visual tips, as in a karaoke). Moreover, it seems it can't do that at FULLSCREEN, which is also required. I know I may create a sketch as big as my screen's resolution, but it's the opposite what I need: change my screen resolution to my application's size. As in old games and such. 

Nevertheless, try it and tell me if you find it useful! Take care, only Autocorrelation works decently, right now. Ignore the other method ;)  

http://playmedusa.com

I was playing around with your code today and  definitely it is much difficult to extract pitch from someone speaking.  As you also mentioned, Processing may not be exactly suitable for your needs. Have you checked openFrameworks and the aubio library? They may be helpful.  

http://www.openframeworks.cc/

http://aubio.org/

On the other hand, my project only requires a person to sing, so I think your code has great potential for it.  I will try to adapt it and let you know how it goes. 

Actually, I am trying to map the pitch of someone singing to the closest semitone. Are there limits for the pitch that can be detected with R2D2? I think it would be enough if I can map a person's singing into a two octaves range with a semitone resolution. Do you think R2D2 could handle this approach with some modifications? I was planning on using logAverages function from Minim library but I am not that sure about how I should proceed, so  any suggestion or idea will be more than welcome :)

Thanks!

Paulo

I'm planning to use Unity3D, the game engine we use for our games, coupled with Processing and minim. These two would capture sound and send it via UDP to a local port. Unity would read from that port and handle the data with C#. That will be ideally faster than Processing. Plus, I have all kind of graphical tools to represent what I need. 

In relation to singing, R2D2 currently handles from 50Hz to 500Hz, or something similar. We usually speak in that range. But we may sing in higher ranges! (440Hz is an A, right?) Anyway, you would better use the Fourier Transform for that, most probably. R2D2 also contains an attempt to use Harmonic Product Spectrum. You could try fixing that, which will probably work nicely for singing! : )

http://playmedusa.com