Let's say that you have a random vector g, then the covariance matrix of gg is defined as K=E{(g−g¯)(g−g¯)†} where the letter E denotes expectation, g¯ denotes the mean of gg, † means transpose for real random vector, and conjugate transpose for complex random vector. The correlation matrix is R=E{gg†} So we have K=R−g¯g¯† For zero-mean random vectors K=R
