C is the answer you get by dong the correlation and lags is the corresponding and what xcorr(A,B) does is really moving A around and see whether it matches up with B.
Let's use a simple example, A = [1 1] and B = [1 1]
In this case, to have non-zero correlation output, A can be moved back and force by one sample and still have some overlap with B.
So at lag = -1, you can imagine that
so C at this point is 1. Similarly for lag = 0, you have
so C at this point is 2. If you do this for every position, you get
C = [1 2 1]
lag = [-1 0 1]
i.e., if the lag is -1 (A move to the left), the correlation is 1, if A does not move, the correlation is 2, ...
From this result, A and B matches best at where the maximum of C occurs, which is lag = 0.
HTH
Best Answer