1
$\begingroup$

What are the fundamental differences between speech and music in the frequency domain?

I can think mainly of features in the time domain that can differentiate speech from music signals, e.g. short silences [Wolfe2002Speech].

Maybe fundamental frequency could be used for distinguishing speech from music? If yes then what would be typical ranges of $F_0$ for speech and music?

$\endgroup$
2
$\begingroup$

One way is to look at the energy below 80 Hz. Unless you have a recording of James Earl Jones, there will be almost none for speech and any music with bass and/or drums is going to have a lot down there.

Note that there is huge variability in music recording and a lot in speech. Creating an algorithm that can reliably distinguish between the two for the majority all recordings is a fairly complicated. Looking at the "average spectrum" isn't that useful since the any individual instance can (and does) look completely different than the "average spectrum". This is actually a good candidate for machine learning although many deterministic algorithms do exist as well.

$\endgroup$
1
  • $\begingroup$ I would think rap music could be especially challenging. $\endgroup$
    – Ed V
    May 26 at 21:08
0
$\begingroup$

(using only the average magnitude spectrum):

  • music often fills in all the spectrum, especially more modern music. Speech is mostly in the mid frequencies

  • in almost all music there is a dominant key. And frequencies at this tone and its overtones will be prevalent enough to cause peaks.

$\endgroup$

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.