Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Generalized cross correlation

31 views
Skip to first unread message

Kamran Iranpour

unread,
Aug 20, 2012, 3:02:59 AM8/20/12
to
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



Bruno Luong

unread,
Aug 20, 2012, 3:19:07 AM8/20/12
to
You should apply GCV on the signal residual, not the signal itself.

Bruno

Kamran Iranpour

unread,
Aug 20, 2012, 7:01:50 AM8/20/12
to
Thanks Bruno, as always right.

Kamran


Bruno Luong skrev:

Kamran Iranpour

unread,
Aug 20, 2012, 7:10:41 AM8/20/12
to
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

Dimitris

unread,
Aug 27, 2012, 6:23:12 AM8/27/12
to
"Bruno Luong" <b.l...@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

ml Lin

unread,
Jan 25, 2013, 11:26:09 PM1/25/13
to
"Dimitris" wrote in message <k1fhqg$hud$1...@newscl01ah.mathworks.com>...
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
0 new messages