What is the most efficient pitch detection algorithms in terms of computation? I need something to implement on a DSP. The typical signal in my situation would be a mixture of low-level wide-band noise plus a series of strong harmonics below 1000 Hz (think of the noise produced by industrial machinery or by car engines). The system needs to sample ("know") the current fundamental frequency of the signal every second or so.


1 Answer 1


YINS fft is the best method that worked for me in the past. I have my open source implementation of it written in awesome D language: link . I compared the results with Praat they match.

  • $\begingroup$ If people put down what is wrong with the post. I can learn what I told is wrong. $\endgroup$ Feb 26, 2016 at 7:47
  • $\begingroup$ i personally found YIN to be far worse than what i could write myself based on the same principles of autocorrelation or AMDF or ASDF. (Octave errors are horrible.) it's an IEEE paper. sorta like that "Constant Q Transform" paper that supposedly looked good on paper but sounded like excrement. dunno who Praat is. $\endgroup$ Feb 26, 2016 at 17:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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