0
$\begingroup$

new to this whole DSP thing. Two related questions:

  • Does there exist an algorithm or classifier that, when given two sound files, would be able to tell me if one is a pitch-shifted version of the other? There are algorithms that can tell you whether two images are resized versions of one another, I feel like this would be the audio equivalent.
  • How would you write an algorithm that takes two recordings that are pitch-shifted versions of one another and spits out the amount of pitch shift. My naive guess would be to count zero-crossings in a fixed window for both recordings and compare the two counts, but I feel like that might lead to aliasing issues.

Thanks!

$\endgroup$

1 Answer 1

1
$\begingroup$

There is an algorithm called Dynamic Time Warping that allows you to measure a distance between 2 time-series indepentently of their relative time stretching. Thus with a correct threshold value, you could detect if the two recording (the original and the pitch-shifted) are actually the same.

Then, by using the warping path, you could determine the relative amount of pitch shift between the 2 recordings.

$\endgroup$

Your Answer

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

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