MATLAB Central Newsreader - xcorr (cross-correlation)

xcorr (cross-correlation)

RV — Tue, 15 Jul 2008 16:47:03 -0400

Hi,

I am using the function xcorr (cross-correlation)as follows:

c = xcorr(x,y,'coeff')

by using 'coeff' it normalizes the sequence so the
autocorrelations at zero lag are identically 1.0.

I would like to calculate the cross-correlation coefficient
between vectors x and y at different lags but without
normalization.

The other 'options'

1) 'biased': Biased estimate of the cross-correlation
function

or

2) 'unbiased': Unbiased estimate of the cross-correlation
function

does not generate an output with the cross-correlation
coefficient.

Thanks for your help

Re: xcorr (cross-correlation)

Andrew Palmer — Tue, 15 Jul 2008 18:49:02 -0400

"RV " <cyanokybus@yahoo.com> wrote in message
<g5ika7$6sg$1@fred.mathworks.com>...
> Hi,
>
> I am using the function xcorr (cross-correlation)as follows:
>
> c = xcorr(x,y,'coeff')
>
> by using 'coeff' it normalizes the sequence so the
> autocorrelations at zero lag are identically 1.0.
>
> I would like to calculate the cross-correlation coefficient
> between vectors x and y at different lags but without
> normalization.
>
> The other 'options'
>
> 1) 'biased': Biased estimate of the cross-correlation
> function
>
> or
>
> 2) 'unbiased': Unbiased estimate of the cross-correlation
> function
>
> does not generate an output with the cross-correlation
> coefficient.
>
>
> Thanks for your help
>
>
>
>
>
>
>
>

I believe there is another option:
3) 'none': use the raw, unscaled cross-correlations

Re: xcorr (cross-correlation)

RV — Mon, 21 Jul 2008 15:59:03 -0400

I guess the question is:

By using c = xcorr(x,y,'coeff')

matlab generates standardize cross-correlation coefficients.
How to generate un-standardize cross-correlation
coefficients (range -1 to 1)?

The other 'options': 'biased', 'unbiased' and 'none' does
not seem to provide cross-correlation coefficients.

Do I need to do an extra calculation to obtain the
correlation coefficient when I ask for the 'none' option?

Thanks

"RV " <cyanokybus@yahoo.com> wrote in message
<g5ika7$6sg$1@fred.mathworks.com>...
> Hi,
>
> I am using the function xcorr (cross-correlation)as follows:
>
> c = xcorr(x,y,'coeff')
>
> by using 'coeff' it normalizes the sequence so the
> autocorrelations at zero lag are identically 1.0.
>
> I would like to calculate the cross-correlation coefficient
> between vectors x and y at different lags but without
> normalization.
>
> The other 'options'
>
> 1) 'biased': Biased estimate of the cross-correlation
> function
>
> or
>
> 2) 'unbiased': Unbiased estimate of the cross-correlation
> function
>
> does not generate an output with the cross-correlation
> coefficient.
>
>
> Thanks for your help
>
>
>
>
>
>
>
>

Re: xcorr (cross-correlation)

Malcolm Lidierth — Mon, 21 Jul 2008 16:37:02 -0400

From the help
'coeff': Normalizes the sequence so the autocorrelations at
zero lag are identically 1.0.

But xcorr does not do that for cross-correlations
so 'coeff' seems to be what you are after if I have
understood. See the scaleXcorr function within xcorr.

It is not unusual to force a correlation of 1.0 at zero lag
in autocorrelations to get rid of rounding errors.