Re: "Pitch detection source code"

On Sep 22, 10:58 pm, Ben Bridgwater <bbridgwa...@xxxxxxxxxxxxx> wrote:
tee_huip...@xxxxxxxxxxxx wrote:
On Sep 19, 7:11 pm, Ben Bridgwater <bbridgwa...@xxxxxxxxxxxxx> wrote:

tee_huip...@xxxxxxxxxxxx wrote:

i still not really understand about the Soft window: hamming,
rectangular,...can anyone explain?

To analyse a signal, i know that i should only use a short-time window/
frame. (20-30ms). What i think is that i just simply pick a range of
data(20-30ms) and do the analysis.
can i do in this way instead of the soft window? If not mistaken there
is some specified algorithm for these soft window that i have to
include them in the Autocorrelation function computation. Is my
statement correct?

Thanks in advance.

D'nise

If you are just going to be doing time domain processing then using a
"range of data" (same as a rectangular window) should be fine. Where
hamming/etc tapered windows are needed is where you are going to be
doing frequency domain analysis (e.g. fourier transform) of the windowed
data and you don't want the window to introduce artificial frequency
components (intoducing a sharp cutoff - rectangular window edge - in the
time domain will introduce a range of high frequency components).

Ben- Hide quoted text -

- Show quoted text -

Hi:

I test my pitch by using a software WASP. what i found out is... when
i pronouce "aaaaaaaa" louder, then i will get my pitch become higher.

is it true that pitch will be affected by loudness of the voice? Are
there any other factors would affect the pitch?

I don't know what degree of correlation between loudness and pitch is
common, but I'd not be surprised if there was at least some (in an
untrained voice), since to increase loudness you're forcing out more air
which requires some increase in muscle tension which may carry over to
the vocal chords.

Other than deliberate pitch control, the main factor affecting pitch is
emotion - stressed/excited = higher pitch, calm/depressed = lower pitch.



For normal pitch, man would have around 125Hz, female 250Hz, Children
300-400Hz. So, what would be the pitch range (for these three groups
trespectively) to be considered as abnormal pitch (vocie disorder)?

I don't think you would classify an unusually high/low voice as a
disorder if the person is healthy and the voice is otherwise normal -
it's just a matter of personal variation. Voice disorders related to the
vocal chords would mostly manifest themselves as changes in quality/mode
of voice rather than just pitch.

Ben- Hide quoted text -

- Show quoted text -

Hi:

i try to use Autocorrelation (AC) function to process my voice signal
before i detect the pitch. After i run my algorithm, i found out one
problem.

first let me explian how i do it.

i only analyse a frame of 600samples (at the centre of my 2 seconds
waveform at 20kHz sampling rate) without using any soft window
technique.

my algorithm for AC function is as below:

/* Calculate the (un-normalized) autocorrelation for a frame of a
signal */

float_vec_t LPCAnalysis::autoCorrelation(const float_vec_t &x)
{
short order=x.size();

float_vec_t R(order);
float sum;
int i,j;

for (i=0;i<order;i++) {
sum=0;
for (j=0;j<order-i;j++) {
sum+=x[j]*x[j+i];
}
R[i]=sum;
}
return R;
}


i expect to get a result in which i should have the largest peak at
i=0. but somehow i couldnt get the largest peak at index=0. Is my
expectation correct? i refer to http://en.wikipedia.org/wiki/Autocorrelation
, in this web i can see the waveform pattern after AC function is
done, the largest peak will be happened on index 0. but why for my
case i couldnt get what i expect?

do the negatives value of waveform affect the result? how if the
negative value of the samples amplitude is larger than the positive
value of the waveform?

thank you....your help is really appreciated!

D'nise

.

Relevant Pages

  • Re: "Pitch detection source code"
    ... is some specified algorithm for these soft window that i have to ... I test my pitch by using a software WASP. ... I don't know what degree of correlation between loudness and pitch is common, but I'd not be surprised if there was at least some (in an untrained voice), since to increase loudness you're forcing out more air which requires some increase in muscle tension which may carry over to the vocal chords. ... disorder if the person is healthy and the voice is otherwise normal - it's just a matter of personal variation. ...
    (comp.speech.research)
  • Re: Robotic voice
    ... The result of an IFFT (in a phase vocoder) is a pulse-like sound ... ... these regular intervals are then the frame length of ... like doing) than the output pitch. ... you don't need an FFT to chop up your voice. ...
    (comp.dsp)
  • Re: Pitch calculation (FFT or autocorrelation)
    ... This is cut-and-paste Matlab routine - just fix the wrapped around ... Pitch is an auditory concept, ... % some parameters (e.g. frame size) will need to be modified. ... % high reference peak magnitude ...
    (comp.dsp)
  • Re: Good voice for bossa nova?
    ... endlessly singing a note and then sliding up or down to where the note ... i think some bossa melodies CAN be complex. ... To me bossa is joao gilberto's voice, vinicius's poetry, and jobim's ... way they treat pitch. ...
    (rec.music.brazilian)
  • Re: AMC: The uptalking? Its got to stop?
    ... I don't know if the actress playing Colby is "uptalking" ... of raising your voice at the end of every sentence. ... in _pitch_, not in volume. ...
    (rec.arts.tv.soaps.abc)