Eigenvalue should be used

[hangtingchen]

Hi, snsun Thanks for your code. I think that we should use the eigenvalue of correlation matrices to solve the permutation problem. https://github.com/snsun/cgmm_mvdr/blob/cfc88aec4e26a6536eb4642566db9701ee9d6369/est_cgmm.m#L63 However, here you used the matrix of eigenvectors. Is this a bug? Another question is that in

MVDR Beamformer Based on Complex Gaussian Mixture Model with Spatial Prior for Noise Robust ASR

The author addressed that

we must normalize the matrices to fairly compare the eigenvalues, while keeping the objective function stable by multiplying the normalization factor by the power parameters.

Do you have any idea of how to realize matrix normalization and keeping objective function stable