Page 1 of 1

AutoRegressive Coefficients - Seem to be large for a normalized signal

Posted: Mon Mar 15, 2021 10:09 am
by openvibeexplorer
Hello there,

I have an EEG signal whose amplitude is normalized between -1 to +1 within each epoch.
There are 256 samples within an epoch.
Each epoch (256 samples) is an input to AutoRegressive box of order 5.
I am getting following auto-regressive coefficients for a single epoch:
1,-4.78718,9.30318,-9.17654,4.59432,-0.933777

The value of these coefficients seem to be large and as I increase the "order" it gets larger.
When I make the order = 15, the coefficients have values in the range of 1000.

Are such high value of AR coefficients expected for a normalized signal?
Or I am doing something wrong?

Edit:
I am working on OpenVibe version 2.2.0

Regards,
openvibeexplorer.

Re: AutoRegressive Coefficients - Seem to be large for a normalized signal

Posted: Sat Mar 20, 2021 6:36 am
by openvibeexplorer
Just to conclude:

I wrote a python script to invoke python module of AR coefficients. And it gives similar results as OpenVibe's AR coefficients.
Which confirms the correct functioning of OV AR coefficients module.

Thanks!
openvibeexplorer.