Axiom-Crux

I used FFFB~ to make a 36 note pitch analyzer, Its working, but i realize that the harmonics of most waveforms are too loud, and it could generally use some refining, unless there's something else that does this that is better, but I couldn't find anything. Ive attached the patch, can anyone think of ways to make this work better? The intent is to use it for an audio visualizer, so it doesn't need to be totally perfect, but hopefully better then this. Or again if anyone knows of an alternative, im open to that.

I used FFFB~ to make a 36 note pitch analyzer, Its working, but i realize that the harmonics of most waveforms are too loud, and it could generally use some refining, unless there's something else that does this that is better, but I couldn't find anything.  Ive attached the patch, can anyone think of ways to make this work better?  The intent is to use it for an audio visualizer, so it doesn't need to be totally perfect, but hopefully better then this.  Or again if anyone knows of an alternative, im open to that.


polyphonic-pitch-detection

made a few small improvements and added the output of the filter to the dac so you can hear, its pretty a wild filter if nothing else.

Hey folks, I know this is an old thread but I am doing work on polyphonic pitch detection for the piano. I was wondering what the state of this research is? Can anyone repost this patch? Or is there something better out there now?

http://cosmal.ucsd.edu/arshia/index.php?n=Main.Transcribe

Completely second that, provided you're dealing with piano. (for those who are coming to this thread at a later date) I believe the implementation is timbre-specific. (It's implemented using Non-Negative Matrix Factorization and uses sonic templates IIRC)

It's CPU hungry but, with the right settings, it works extremely well on piano.

Avoid the filter based approach; it will not work nearly as well, and will have problems with low notes. (The lowest notes on the piano have harmonics that can be louder than the fundamental)

Completely second that, provided you're dealing with piano.  (for those who are coming to this thread at a later date)  I believe the implementation is timbre-specific.  (It's implemented using Non-Negative Matrix Factorization and uses sonic templates IIRC)

Avoid the filter based approach; it will not work nearly as well, and will have problems with low notes.  (The lowest notes on the piano have harmonics that can be louder than the fundamental)


Polyphonic pitch detection