PureBasic Forums - English

I am looking for a pb autocorrelation routine.
I have a row of 400000 numbers, and i like to do an autocorrelation on it.
( For my SETI - program , just to find is there a system in the noise i do receive....)

I can find enough mathematics for it... but i can not transform that to a pb program.

Thanks Wim

Maybe using Wiener's method of auto correlation using an FFT / IFFT

Idle: yes i do read information about FFT and IFFT, but there has to be something in
between.. but i do not understand what.... (i have routines for a FFT and an IFFT)


Ideasvacuum: Your tips give nothing about autocorrelation... (if i am right)
but there was something like rosettacode.org... that looks great. but nothing
about autocorrelation....


By the way: good to see you both again, and thanks for your help !!!!!!

idle, should it be as follows?

you do a FFT on the row of numbers.
You get then two rows of numbers... the reel part and the imaginair part. REX(i) and IMX(i)

if i do understand the wiener method i have to multiply the reel part with the imaginair part
This gives a new row of numbers.

Then i have to do an IFFT on that row of numbers....

Should this be it?????


.

Hi Wim -well essentially, autocorrelation is crudely based on statistics. Have a look at the subject on Wikipedia.

Yes, ideasVacuum,
i have looked at wikipedia, and lots of other sites.. but i can find only unreadable mathematical formules...

This is always my problem...

Wim

Well, this topic is not simple.
Here is an example lecture on convolution...
Once you know the theory, then comes the application in Digital Signal Processing or Statistics or Astronomy or whatever field you are pursuing.

Hi Skywalk,
It is a very nice lecture. Nice to follow, but convolution is not what i need.
A lecture about autocorrelation would be fantastic.

Wim

Hi wimapon,

Have you tried googling 'autocorrelation algorithm' ?

You might, perchance, find something there.

Regards

Dave

Yes Davido, i did that.
I come everytime on the same things.... or unreadable formulas of using of mathlab....

Wim

Numerical Recipes has code to do autocorrelation, it's available in Fortran, C and Pascal.

jack, ik heb ze gevonden, maar ik kan er niets mee......
ik kan de codes niet naar basic overschrijven...
jammer