I have a Mixture of Gaussians to model an arbitrary distribution. I would like to model a distribution derived from this GMM with:

Mean = Weighted average mean of GMM means.

I am not sure about how to combine the co variances of each clusters into one. Is there a standard way for doing this?