About the normalization in the modified k-means algorithm

[neurothew]

Let me post the issue here...not sure if anyone will see it.

In modkmeans.m, in step 2a of the function opt_segment, there is a step of normalization.

A = bsxfun(@rdivide,A,sqrt(diag(A*A')))

However, I am confused by the operation right here, which is also not clearly written in the paper. I am guessing it is trying to calculate the covariance matrix and extracting the diagonal element (which is then the variance of each channel). But there are 2 questions:

1. The covariance can be calculated via AA' only when the data is centered to mean 0 prior to the calculation (but I did not see any of such centering operation)
2. There should be a factor of 1/n-1, where n is the number of data points.

am i missing anything? any help is much appreciated, thanks!