I know that crosscorr function normalizes the correlation and gives us correlation with respect to 1. Where as the xcorr function of matlab gives unnormalized values. But why I am not able to see any correlation with crosscorr function when both xcorr and crosscorr are supposed to indicate correlation between two signal.
Now I attach three signal A , B and C . A and B are similar, and C is noise. So Iam supposed to get good correlation between A and B and least with C.
Using xcorr Iam able to see this clearly but crosscorr doen't tell anything. Why?????????????????????????????????
you are correct that crosscorr normalize the correlation. "Normalize" here however, means that the signals will be "detrended" first (substracted by its mean) and then divided by its standard deviation.
I was able to create a same output doing:
crosscorr(a,b)
compared to:
1/(n-1) * xcorr(b-mean(b),a-mean(a))/(std(a)*std(b))
with n is the number of elements.
Another thing that is different is that in using xcorr, the shorter data will be appended by zeros so that the data has the same length.
Usually, when you have a triangular shape for the autocorrelation function, it means that you are using the 'biased' estimator instead of the 'unbiased' one. So it is better to use the following Matlab command, it will be clearer :
>> [Corr_function,Lags] = xcorr(A,B,''unbiased');
>> plot(Lags, Corr_function)
There was a typo into my previous matlab code :
>> [Corr_function,Lags] = xcorr(A,B,'unbiased');
is better :-)
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software