cross correlation

[Fabrice François]

Hello
I have two temporal signals that I want to cross-correlate each other.
I use a sliding window of a finite length that I apply to each signals and
I calculate the cross-correlation betwwen the two signals over this window
(over this window, I also apply one smoothing ie Blackman or Hamming window)
..
But when this windows arrives at the end of my temporal signal, the value of
the calculated cross correlation is of non-sense (by plotting the two
signals, it is possible to calculate by the hand a realistic value of the
cross correlation).
Does anyone know why ?

Extra question : it seems necessary to use a normalization of the input
signal before usin the vi Cross-correlation. Is it true and why
Thanks in advance
Fabrice

[Eric Hervieu]

Hi Frenchie

Verify your local variables and it will go better...

Ah Ah Ah !!!


"Fabrice François" a écrit dans le message de
news: amvbmk$j7g$1@ellebore.saclay.cea.fr...
> Hello
> I have two temporal signals that I want to cross-correlate each other.
> I use a sliding window of a finite length that I apply to each signals
and
> I calculate the cross-correlation betwwen the two signals over this window
> (over this window, I also apply one smoothing ie Blackman or Hamming
window)
> .
> But when this windows arrives at the end of my temporal signal, the value
of
> the calculated cross correlation is of non-sense (by plotting the two
> signals, it is possible to calculate by the hand a realistic value of the
> cross correlation).
> Do
es anyone know why ?
>
> Extra question : it seems necessary to use a normalization of the input
> signal before usin the vi Cross-correlation. Is it true and why
> Thanks in advance
> Fabrice
>
>

[Andrew Madry]

Regarding the normalisation, the cross correlation vi provides an unnormalised result (see the help for the equation)
To normalise you can calculate the rms of each of the input arrays, multiply these and multiply this result by N, the size of the input array. To normalise the cross correlation divide the result of the cross correlation vi by this number. When the function is normalised the result will always be 1 at zero lag.

Regarding the first part of the question I'm not sure I understand. You should be able to window the input arrays before cross correlating. See the book by Bendat and Piersol: "Engineering applications of correlation analysis" for the implications of this.