cumulants

[xDR1TeK]

Hello all, Please help.

I was working on a way to complete an algorithm for automatic modulation classification by extracting feature from the Spectral Coherence function and the cyclic domain profile.

Found out that feature extraction by cyclo-stationary analysis fairs poorly for balanced modulation schemes like QAM and everything quadrature.

So the other step for feature extraction is by finding cumulants.

Is there a SubVI that does that? The math is too advanced for my simple mind.

I have found a SubVI called Continuous Cumulative Distribution Function and another for the moments.

My problem is that I need to generate cumulants of moments for an incoming signal, that is made of an array of 20,000 elements.

How to get the moments and cumulants from an array that represents an incoming Signal?

It is far easier said than done. The math is not clear for direct application yet i am getting to terms with the concept behind it.

As I mentioned, I have no idea how to add this in LabVIEW.

This is a table of the Cumulants generated, retrieved from :

Blind Signal Identification for Emerging Intelligent Radio Systems: Classical Approaches and New Trends by Octavia A. Dobre, Memorial University, Canada

The equations below from http://mathworld.wolfram.com/Cumulant.html

and i can't make sense of them, this is quite a complex topic, and I can't add everything I have learned about cumulants and statistical approach for ergodic signals here.

So I hope you believe I have some understanding of the topic and advise me in that respect.

This is for an FYP using USRP. I think i bit off far more than I can chew.

[aputman]

Sorry for giving you the false impression that someone actually had an answer.  I know you are speaking words and they are some really good words but all I hear is "blah blah blah blah".

GOOD LUCK

(I did use a lot of smilies in this post but they're so awesome. )

[Hatef.fouladi]

maybe it could help you
http://zone.ni.com/reference/en-XX/help/371419D-01/lvtimeseriestk/tsa_exponential_average/

[Hatef.fouladi]

there is some example of labview for that too

[Shayban]

Had to post through this account, the one is not registering comments as my nick, it is reverting to this one, so please be patient with this buggy site and nick switch.

[Shayban]

Hello again, thanks for the effort.

You are walking in my footsteps, you are almost there.

I already tried it, there is no way to control moment generation into cumulants in that subvi.

However, I feel i got some breakthrough, and will discuss it here soon enough.

There is a way to get the moment of random values from an array, the array that is the stream of data coming from my signal generator.

I used Moment about Mean VI and then used the expanded Maclaurin Series (that i didn't know how to expand since the math is too literate) to calculate the moments.

Now there was a trick, univariate cumulants generated somewhat standard results and matched the multivariate cumulants in certain measurements.

But still, I needed to find a way to generate the multivariate cumulants from the moments.

Luckily, I found the steps of finding the multivariate cumulants by setting the order of moments and combining them in a mathematical equation that includes summation, absolute value, and conjugation of the input signal array.

Need to format the ideas better and repost here to get some more feedback on feasibility.

and now i am getting posting errors.mmmmmm