Covariance matrix Kmeans in matlab

Hi, I would like to generate the covariance matrix using kmeans algorithm in Matlab as initial parameter estimation. Later on, i want to use the parameter estimation in gaussian mixture model.

This is the code in matlab:

[idx,C] = kmeans(Faithful,2);
theta.mu = C';
theta.tau = hist(idx,2)/length(idx);
for j = 1:K
        theta.Sigma(:,:,j) = cov(Faithful(idx==j,:));
end

Is that the correct way to estimate the covariance matrix? The result yielded:

(:,:,1) =
    0.1722    0.7471
    0.7471   31.7966
(:,:,2) =
    0.1543    1.0129
    1.0129   35.5095

This is because, when i checked from mixtools package in R, the results are bit different whereas the covariance elements in second covariance matrix are more than 1 as shown in above compared to the answer below:

[[1]]
           [,1]       [,2]
[1,] 0.06916788  0.4351698
[2,] 0.43516975 33.6972966

[[2]]
          [,1]       [,2]
[1,] 0.1699681  0.9406057
[2,] 0.9406057 36.0461700

If i want to compute the pdf, can i simply recall the parameter estimate for theta.Sigma like this?:

densEst(pointsObserved,j) = mvnpdf(X(pointsObserved,:),theta.mu(:,j)',theta.Sigma(:,:,j));