Suppose I have some time series $s(t)$ which contains Gaussian white noise generated by a distribution $N(0,\sigma^2)$

Then I apply a filter to s(t) with a frequency response $H(\omega)$, giving me $s_H(t)$

What distribution does $s_H(t)$ have?

I'm particularly interested in cases where $s_H(t)$ is band limited Gaussian white noise.

  • $\begingroup$ This question seems to be almost exactly same as the one I'm asking here but sadly, it has no answer: dsp.stackexchange.com/questions/38428/… $\endgroup$ – alessandro Dec 31 '17 at 3:01
  • $\begingroup$ If the answer below answered your question you can accept it by clicking on the check mark to its left, thanks! $\endgroup$ – Matt L. Jan 1 '18 at 11:12

If you filter a Gaussian random process with an LTI system, the output will also be Gaussian. You can make intuitive sense of this by considering that a linear combination (which is what filtering does) of jointly Gaussian random variables is a Gaussian random variable.

You can find an in-depth treatment of filtering random processes in this MIT OpenCourseWare document (section 7.4).

| improve this answer | |
  • $\begingroup$ Reworded second sentence. Please put back if you disagree. $\endgroup$ – Dilip Sarwate Dec 31 '17 at 15:10

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.