Eigenvectors for demixing matrix initialization in AuxIVA or OverIVA

[aanugraha]

Hi!

I'm curious about the use of eigenvectors for initializing the demixing matrix in AuxIVA or OverIVA.
To be specific, I have a concern regarding line 169, as shown in the following code snippet.

pyroomacoustics/pyroomacoustics/bss/auxiva.py

Lines 163 to 169 in 3ea2abb

	if init_eig:
	# Initialize the demixing matrices with the principal
	# eigenvectors of the input covariance
	v, w = np.linalg.eig(Cx)
	for f in range(n_freq):
	ind = np.argsort(v[f])[-n_src:]
	W[f, :, :] = np.conj(w[f][:, ind])

I think a transposition is missing and the line should be:

W[f, :, :] = np.conj(w[f][:, ind]).T

Using the current line of code, we face a shape mismatch problem for OverIVA but not for AuxIVA, which is misleading.

CMIIW. Thanks!