Pitch tracking techniques
Auto-correlation: if there's a basically repeating waveform, computer can detect when the differences between samples and past samples are minimal (when it's basically repeating), and so computer can figure out period of repeating sound, will use that period to calculate a frequency--not too much worry about time delay, probably how guitar tuners work--not necessarily fast enough for performance
Spectral analysis: fiddle~ and sigmund~ use this; finds harmonic relationships between peaks in frequency spectrum, uses that to determine fundamental frequency
Sigmund is best at outputting note and volume information, with arguments "notes" and "env." What you do with that information is up to you.
It misses notes if you play fast, and it doesn't like multiple pitches at once.
If there's a danger of it picking up its own sound or an ensemble member's sound, Chris measures threshold (sound of audience, other players), and sets it so that sigmund~ only pays attention to sound above that threshold. It's also a good idea to set sigmund to ignore sounds that are below the natural range of your instrument.
peakamp~ object: reports loudest amplitude in amount of time you set--can set it so that if peak amplitude is below a certain level, ignore it
*on Thursday, stay tuned for talking about transfer functions for mapping domains onto a third, "bridging" domain
No comments:
Post a Comment