0
$\begingroup$

I'm reading the paper Model-Driven Deep Learning for Joint MIMO Channel Estimation and Signal Detection by Hengtao He, Chao-Kai Wen, Shi Jin, and Geoffrey Ye Li on Orthogonal Approximate Message Passing (OAMP) detectors.

There is an step that it is getting me crazy, in Equation $(42)$ in Appendix B. It is the following:

enter image description here

At first, I thought that it was a consequence of a Hypothesis, but then, to derive the next expression in Equation $(44)$, it is used again this formula.

enter image description here

Is there some kind of property which I might be missing?

Regards.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

I am not really sure which part you didn't understand. I'm assuming it's the step that goes from the rightmost expression at the top to the one at the bottom. Assuming $\mathbf{q} = \mathbf{x} - \hat{\mathbf{x}}$ is i.i.d. Let me give it a try...

$ \begin{aligned} \textrm{E}[ (\mathbf{x} - \hat{\mathbf{x}})^{\mathrm{H}} \mathbf{H}^{\mathrm{H}}\mathbf{H} (\mathbf{x} - \hat{\mathbf{x}}) ] & = \textrm{tr} \{ \textrm{E}[ (\mathbf{x} - \hat{\mathbf{x}})^{\mathrm{H}} \mathbf{H}^{\mathrm{H}}\mathbf{H} (\mathbf{x} - \hat{\mathbf{x}}) ] \} \\ & = \textrm{E}[ \textrm{tr} \{(\mathbf{x} - \hat{\mathbf{x}})^{\mathrm{H}} \mathbf{H}^{\mathrm{H}}\mathbf{H} (\mathbf{x} - \hat{\mathbf{x}}) \} ] \\ & = \textrm{E}[ \textrm{tr} \{\mathbf{H}^{\mathrm{H}}\mathbf{H} (\mathbf{x} - \hat{\mathbf{x}}) (\mathbf{x} - \hat{\mathbf{x}})^{\mathrm{H}} \} ] \\ & = \textrm{tr} \{\mathbf{H}^{\mathrm{H}}\mathbf{H} ~\textrm{E}[ (\mathbf{x} - \hat{\mathbf{x}}) (\mathbf{x} - \hat{\mathbf{x}})^{\mathrm{H}} ] \} \\ & = \textrm{tr} \{\mathbf{H}^{\mathrm{H}}\mathbf{H} ~\textrm{E}[ (\mathbf{x} - \hat{\mathbf{x}}) (\mathbf{x} - \hat{\mathbf{x}})^{\mathrm{H}} ] \} \\ & = \textrm{tr} \{\mathbf{H}^{\mathrm{H}}\mathbf{H} ~\sigma_{q}\mathbf{I}] \} \\ & = \sigma_{q} \textrm{tr} \{\mathbf{H}^{\mathrm{H}}\mathbf{H}] \}. \end{aligned}$

where I am assuming $\sigma_{q} = \textrm{E}[ \|(\mathbf{x} - \hat{\mathbf{x}}) \|^2 ] = \frac{\textrm{E}[ \|\mathbf{q} \|^2 ]}{N_t} $.

As you can see, I am making a lot of assumptions here. I don't know if they hold since I couldn't access the paper. But I think it can shed some light on your problem (if your question is strictly about the trace operations).

Improve this answer
$\endgroup$
1
  • 2
    $\begingroup$ Yes, you are right. I was so focus in the techincality and formulas that I was missing the i.i.d hypothesis. Thanks! $\endgroup$
    – Cobretti
    Jan 9, 2021 at 15:19

Your Answer

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

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