if $x(t)=b A e^{ j\omega t} + e(t)$ for $t= 1,2,...,N$ where $b$ is a parameter, $A$ is a vector $M \times 1$, $e(t)$ is a white Gaussian noise with covariance matrix of $Q$ theh what is log-likelihood function for that?

  • $\begingroup$ why is it : $log|Q|+Trace[Q^{-1}*C]$ ; where $ C=1/N * summation((x-bAexp(jwt)(x-bA*exp(jwt))^{H}) $ $\endgroup$ – Reza Mahjoob Sep 30 '17 at 19:58
  • 1
    $\begingroup$ Hi Reza, instead of a comment, you can edit your question. $\endgroup$ – Atul Ingle Sep 30 '17 at 20:09

I’m assuming that the trace is what is confusing, so let me give you some clues.

First an identity $$ Trace(\mathbf{ABC)})= Trace(\mathbf{BCA}) $$

and for scaler $x$, $$ Trace(x) =x$$

Let $\mathbf{a}$ and $\mathbf{b}$ be $N \times 1$ vectors and $\mathbf{C} $is a $N \times N$ matrix, so $$ \mathbf{ a^T C b} = Trace ( \mathbf {C \,b a^T}) $$

The rest you should be able to figure out using the definition of a Gaussian, independence, and properties of the log.

  • $\begingroup$ Thanks a lot. What is the need for "abs" of "Q" in "log |Q| "? $\endgroup$ – Reza Mahjoob Oct 3 '17 at 18:22
  • $\begingroup$ not abs, determinant $\endgroup$ – Stanley Pawlukiewicz Oct 3 '17 at 18:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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