Question on using EM algorithm for linear regression and GMM assumption.

I have a data set which are a mixture of K different Gaussian components. They consist of input data $X \in R^{mxN}$ and $Y \in R^{nxN}$. I have to find the regression coefficient (M) for each components, where Y = M * X + E, E is the noise part. Also I have to calculate variance of E. I googled that but no success! I am strongly looking for some helps or references.