smoothing

Omega · Post by **Omega** » Tue Oct 30, 2007 11:58 am

Suppose I have a set of points as shown below

1

Baltazar · Post by **Baltazar** » Tue Oct 30, 2007 12:05 pm

any low pass filtering will "smooth" your serie. Moving average being one of them.

Do you have any other requirements? (like causality for example, compact spectrum...)

If you have matlab, thank to fourier transform, you can do it very quick

FDAXHunter · Post by **FDAXHunter** » Tue Oct 30, 2007 12:47 pm

I think he's looking for a non-causal filter, because there does seem to be a forward looking / neighbor effect that he desires. But a centered moving average or any other smoothing method will do, like Baltazar said.

Omega · Post by **Omega** » Tue Oct 30, 2007 1:03 pm

Cheers guys. For the moment I'm just looking for something basic. Cheers. I'll look into that avenue. My only one condiiton is that the sum of the two series should be the same

Baltazar · Post by **Baltazar** » Tue Oct 30, 2007 1:59 pm

then that mean that your smoothing element should be of integral 1

In your case the smoothing element is just [1/2 1/2] ie you do an average of two points. You could use [1/3 1/3 1/3], a gaussian, hamming, haning, blackman,some exponential, stuff like this [1/8 2/8 5/8 0].

To be sure your filter is causal, it has to be filter(x)=0 for x>0 (it actually depends wich convolutio operator you use, some convolve with the reversed filter).

Maybe all of this does not help, then just look for casual low pass filtering, it should cover your need.