Thread Subject:

Generalized cross correlation

Generelized cross-correlation (GCC) is said to be a better estimate of
cross-correlation for time-delay estimation between channels. It is a
whitening technique on cross-cspectra.
I have this small code for computing the GCC:

fChannel1 = fft(channel1);
fChannel2 = fft(channel2);
conjFChannel2 = conj(fChannel2);
num = (fChannel1).*conjFChannel2;
den = abs((fChannel1).*(conjFChannel2));
fGeneralized = num./(den+eps);
generalizedCrossCorrelation = ifft(fGeneralized);
[maxcorr phind] = max(fftshift(generalizedCrossCorrelation));

But my "phind" which is is the index of maximum is always zero. Is
anyone familiar with the technique?

Thanks,

Kamran

You should apply GCV on the signal residual, not the signal itself.

Bruno

Thanks Bruno, as always right.

Kamran


Bruno Luong skrev:
> You should apply GCV on the signal residual, not the signal itself.
>
> Bruno

On 20 Aug, 09:19, "Bruno Luong" <b.lu...@fogale.findmycountry> wrote:
> You should apply GCV on the signal residual, not the signal itself.
>
> Bruno

Thanks Bruno, as always correct.

Kamran

"Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <k0sodb$ps$1@newscl01ah.mathworks.com>...
> You should apply GCV on the signal residual, not the signal itself.
>
> Bruno

Hi Bruno,
can you be more specific?

Dimitris

"Dimitris" wrote in message <k1fhqg$hud$1@newscl01ah.mathworks.com>...
> "Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <k0sodb$ps$1@newscl01ah.mathworks.com>...
> > You should apply GCV on the signal residual, not the signal itself.
> >
> > Bruno
>
> Hi Bruno,
> can you be more specific?
>
> Dimitris

Hi Bruno,

I am encountering this problem too.
Do you think you can provide us with more details?
Do you mean that I need to create a specific model to get the residual signal?
Or perhaps you can make slight amendments to Kamran's original code to enlighten us?

ML