Can the XCOV function return a result that is both unbiased and normalized?

Question

The XCOV function in the Signal Processing Toolbox 6.8 (R2007b) has four distinct options: 'biased', 'unbiased', 'coeff', and 'none'. I would like to be able to create unbiased estimates of the cross-covariance function of my sequence with a normalized zero lag term.

John Michell · Accepted Answer

The ability to create estimates of the cross-covariance function that are both unbiased and normalized is not available in the Signal Processing Toolbox.
 
To work around this issue, create unbiased estimates and normalize by the zero lag term manually, as in the following example:
 
ww = randn(1000,1);

N = 10;

[cov_ww,lags] = xcov(ww,N,'unbiased');

cov_ww = cov_ww / cov_ww(N+1);

stem(lags,cov_ww)

Can the XCOV function return a result that is both unbiased and normalized?

Get a Free Consultation or a Sample Assignment Review!