What should the energy, pitch frequency, and zero crossing rate of a voice/unvoiced/silence frame in signal be?

My professor said that it is experimental, but I do not have any idea what the factors of these are and where I should begin from.

I would appreciate any guesses based on experimental data or any limits on the quantities.

I've tried a zero crossing rate between 2000 and 94000, energy between 0.0001 and 0.7, and pitch frequency between 180 and 500.


Are you being asked to write a simple vocoder?

Your question is somewhat vague and confusing, but it seems like you are being asked to determine parameter thresholds for deciding when a speech frame is voiced (has fundamental periodic component and harmonics), unvoiced or frictional (noise generated), and silent (do not encode).

It looks like you would be using the pitch frequency estimates to determine if the frame is voiced/unvoiced, and signal energy to determine if it is a silence frame.

If he said it's experimental, I think he is asking you to look at the energy and pitch content of an audio signal and see if you can see differences that will help you choose thresholds. For example, get some audio of someone speaking. When the person is vocalizing (i.e. vowel sounds like "aaaahh" and nasally sounds like "mmmmmmmm") the spectrum should show periodic components. For unvoiced sounds (i.e. short "th" "t" "s" "f" type sounds) the spectrum will look more like noise. And when there is silence, the signal should be relatively low energy. Take what you find and use that information to determine your thresholds.

  • $\begingroup$ Thanks alot,and yes the point is about determining thresholds... $\endgroup$ – drizzle Jan 17 '14 at 19:48

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.