I'm quite interested in real time pitch tracking. I have used mcleod pitch method, Yin and other methods to perform real time pitch tracking for a monophonic audio track.

I know that Polyphonic pitch detection is at a whole different level and I want to know where to start. If the experts(masters) could point me at the right direction, It would be greatly appreciated. Papers, books, Topics or any suggestions that you could give me would be of great help.

  • 1
    $\begingroup$ Peter Neubäcker is not, as far as i know, on this Stack Exchange (or any other). dunno what other "expert" or "master" to refer to. polyphonic pitch detection is (a female canine and is) pretty closely related to, what we sometimes call source separation. best guess i can give you is use an autocorrelation-like detector and pick the strongest peak, then tune a comb filter tuned to that same fundamental frequency to remove the note that corresponds to that strongest peak. then do autocorrelation again on the residual remainder. repeat the whole thing until the rest sounds like mush. $\endgroup$ – robert bristow-johnson Jun 8 '18 at 9:57
  • $\begingroup$ Thank you Mr. johnson, I'll dig into source separation :) $\endgroup$ – Jedi Nerd Jun 8 '18 at 10:13

It's a research topic. So try the music-ir.org MIREX web site for a vast number of academic papers and transcription competition entries.

If you want to try a newer direction, try gathering many 100's of hours of snippets of tagged music, fuzz it, and feeding that as training data to a DNN that includes convolutional+LSTM layers.

  • $\begingroup$ ya know, i just forgot about MIREX. that is absolutely the best answer about "where to [go]". $\endgroup$ – robert bristow-johnson Jun 9 '18 at 3:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.