I have a vector containing all the PSD of a series of signals

$\underline{s} = \left[ |S_{x_1x_1}| |S_{x_2x_2}| |S_{x_3x_3}| ...\right]^T$

I also have a matrix $\gamma^2$ containing all the coherences

$\gamma_{ij} = \frac{|S_{x_ix_j}|^2}{S_{x_ix_i} S_{x_jx_j}}$

I would like to obtain the matrix of the cross-spectral densities, i.e. a matrix $\mathbf{S}$ so that $S_{ij} = |S_{x_ix_j}|^2$

Is it possibile to obtain it from $\underline{s}$ and $\mathbf{\gamma}$ using matrix notation?

The only way I found at the moment is:

$\mathbf{S} = (s s^T) \circ \gamma$

where $\circ$ indicates the Hadamard (element-wise) power and multiplication. This is however seldom used and usually element-wise multiplication is avoided in linear algebra.

Is there another way to do it?


1 Answer 1


$S= diag(\underline{s}) \cdot \underline{\underline{\gamma}} \cdot diag(\underline{s})$

  • $\begingroup$ What does double underline stands for? $\endgroup$ Commented Jun 5, 2022 at 6:14
  • $\begingroup$ @GideonGenadiKogan indicates that $\gamma$ is a matrix. (after reading this probably this wasn't the best choice... math.stackexchange.com/questions/3653351/…) $\endgroup$
    – Luca
    Commented Jun 10, 2022 at 11:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.