Re: How formants structures affect autocorrelation peaks detection?

On Oct 1, 3:39 pm, Tony Robinson <to...@xxxxxxxxxxxxxxxxxxxxxxxx>
wrote:
tee_huip...@xxxxxxxxxxxx writes:
On Sep 30, 5:17 pm, Tony Robinson <to...@xxxxxxxxxxxxxxxxxxxxxxxx>
wrote:

Did you have a look
the spectrogram of some voiced speech as I suggested?

i did. But sorry to tell you that i am not able to understand the
spectogram, the relationship between energy, formant, and the
fundamental frequency - pitch.

Fine. Did you see the pitch as regular repeats of the waveform in the
time domain and as harmomics in the frequency domain? You may need to
set your spectrogram to have a longer window length (say ~50ms) in order
to see the pitch in the frequency domain.

Tony

The major Peak in the autocorrelation function corresponds to pitch
harmonics (regularly spaced minor peaks) in the (log) power spectrum,
as Tony mentioned. Separation between formants is far greater, in
general, than the regular separation between pitch harmonics in the
spectrum. Moreover, formants are not regularly placed in the spectrum.
So, peaks (if any) in the autocorrelation function due to formants are
really small compared to those of pitch harmonics. Thus, formants are
unlikely to affect pitch peak detection in AC function.

-chief

.

Relevant Pages

  • Re: How formants structures affect autocorrelation peaks detection?
    ... harmonics in the power spectrum, ... formants are not regularly placed in the spectrum. ... peaks in the autocorrelation function due to formants are ... really small compared to those of pitch harmonics. ...
    (comp.speech.research)
  • Re: Logarithmic Interpolation
    ... interpolate the FFT logarithmically, shift the contents towards ... like interpolating the signal in the time-domain? ... Interpolation (i.e. sampling the spectrum on a denser grid) ... pitch shift, which is multiplying every frequency by a certain ratio. ...
    (comp.dsp)
  • Re: DFT and resonance -- question
    ... today for simulating how the ear responds to the frequency spectrum? ... producing a sharp visual spectral representation of a sound would be to ... In other words, if you want a frequency resolution, df, of 0.1 Hz, ... precise pitch analysis I was interested in. ...
    (comp.dsp)
  • =?ISO-8859-1?Q?Re=3A_Redefining_=5Ba=5D_and_=5B=E6=5D=3F_or_using_a_better_vowel?= =?ISO-
    ... F0 (pitch) between the following two lines in a hymn might have the ... F1 and F2 are resonance frequencies, based on the shape of the mouth ... Of course, there will be minor variations in the formants, because ... (I even have a tape of a singer singing the five vowels at frequencies ...
    (sci.lang)
  • Re: DFT and resonance -- question
    ... producing a sharp visual spectral representation of a sound would be to ... the resulting spectrum is quite blurry ... precise pitch analysis I was interested in. ... not chainsaws... ...
    (comp.dsp)