I'am studying robust methods for processing power electrical signals in non-Gaussian noise. I know the mostly used method is M-Estimator and have already implemented an algorithm to analyze power harmonics. But how can I evaluate it and compare it with other non-robust methods. Besides impulsive noise, is there any other non-Gaussian noise? How can I simulate them? Thanks in advance!!!!!
How can I simulate non-Gaussian noise?
Started by ●December 30, 2008
Reply by ●December 30, 20082008-12-30
>I'am studying robust methods for processing power electrical signals >in non-Gaussian noise. I know the mostly used method is M-Estimator >and have already implemented an algorithm to analyze power harmonics. >But how can I evaluate it and compare it with other non-robust >methods. Besides impulsive noise, is there any other non-Gaussian >noise? How can I simulate them? Thanks in advance!!!!!There is no bound to the kinds of non-Gaussian noise that you can simulate. Any noise with a pdf other than Gaussian is by definition non-Gaussian noise. You should start with what your application is, and see if there is a good model for the kind of noise present in *your* measurements, even if you are going to do a simulation of it. Here are a few models of practical importance: http://www.tutorialsweb.com/rf-measurements/noise-figure/types-of-noise.htm What is your application? Are you trying to solve a "real" problem? Emre
Reply by ●December 30, 20082008-12-30
On Dec 30, 8:53�pm, jjdoc <tsaitow...@sina.com> wrote:> I'am studying �robust methods for processing �power electrical signals > in non-Gaussian noise. I know the mostly used method is M-Estimator > and have already implemented an algorithm to analyze power harmonics. > But how can I evaluate it and compare it �with other non-robust > methods. Besides impulsive noise, is there �any other non-Gaussian > noise? How can I simulate them? �Thanks in advance!!!!!I think you are confusing the stationary distribution of a stochastic process, with the color of the process. For example, you can make an impulsive noise with ANY distribution, since the two characteristics are independent of each other. Maybe you can elaborate on the exact problem statement that you are assigned?
Reply by ●December 30, 20082008-12-30
On Dec 31, 10:13�am, "emre" <egu...@ece.neu.edu> wrote:> >I'am studying �robust methods for processing �power electrical signals > >in non-Gaussian noise. I know the mostly used method is M-Estimator > >and have already implemented an algorithm to analyze power harmonics. > >But how can I evaluate it and compare it �with other non-robust > >methods. Besides impulsive noise, is there �any other non-Gaussian > >noise? How can I simulate them? �Thanks in advance!!!!! > > There is no bound to the kinds of non-Gaussian noise that you can > simulate. �Any noise with a pdf other than Gaussian is by definition > non-Gaussian noise. �You should start with what your application is, and > see if there is a good model for the kind of noise present in *your* > measurements, even if you are going to do a simulation of it. �Here are a > few models of practical importance:http://www.tutorialsweb.com/rf-measurements/noise-figure/types-of-noi... > > What is your application? �Are you trying to solve a "real" problem? > > EmreThany you, Emre. I've read that webpage, but I still don't know how to simulte them. If I want to simulate impulsive noise, how to write a matlab program?
Reply by ●December 31, 20082008-12-31
>Thany you, Emre. I've read that webpage, but I still don't know how to >simulte them. >If I want to simulate impulsive noise, how to write a matlab program? >See below link: http://en.wikipedia.org/wiki/Shot_noise#In_electronic_devices You need to generate a Poisson random process. If you have access to Statistics toolbox for Matlab, "help random" in the command window. Otherwise you may write your own code. (A quick google search should turn up a lot on how to do that.) But more importantly you should first understand what your problem really calls for. Try avoiding "a shot in the dark." Emre
Reply by ●December 31, 20082008-12-31
On Dec 30, 10:11�pm, jjdoc <tsaitow...@sina.com> wrote:> On Dec 31, 10:13�am, "emre" <egu...@ece.neu.edu> wrote: > > > > > >I'am studying �robust methods for processing �power electrical signals > > >in non-Gaussian noise. I know the mostly used method is M-Estimator > > >and have already implemented an algorithm to analyze power harmonics. > > >But how can I evaluate it and compare it �with other non-robust > > >methods. Besides impulsive noise, is there �any other non-Gaussian > > >noise? How can I simulate them? �Thanks in advance!!!!! > > > There is no bound to the kinds of non-Gaussian noise that you can > > simulate. �Any noise with a pdf other than Gaussian is by definition > > non-Gaussian noise. �You should start with what your application is, and > > see if there is a good model for the kind of noise present in *your* > > measurements, even if you are going to do a simulation of it. �Here are a > > few models of practical importance:http://www.tutorialsweb.com/rf-measurements/noise-figure/types-of-noi... > > > What is your application? �Are you trying to solve a "real" problem? > > > Emre > > Thany you, Emre. I've read that webpage, but I still don't know how to > simulte them. > If I want to simulate impulsive noise, how to write a matlab program?I agree that defining the model is key. Once you have it, the m-file "starnd.m", which you can find online, might be useful. http://www.tsi.enst.fr/icacentral/icalistArchive/2004/0642.html John