Page 1 of 1

What being fed into the SWLDA

Posted: 12 Feb 2020, 21:05
by sophie
Hello all:

I am very new to the P300 speller. As I use only the GUIs to analyze data, but what are the features fed to SWLDA? Is it raw continuous data from all 8 channels (I use 8-electrode cap), or filtered data? is it downsampled or sampled at 256 hz? where do I find this information?

Thank you!

Re: What being fed into the SWLDA

Posted: 13 Feb 2020, 08:15
by mellinger
Hi,

you will find more details about P300 classification in this paper:
https://hal.archives-ouvertes.fr/file/i ... e_2006.pdf
Preprocessing is described in section 2.4.

HTH,
Juergen

Re: What being fed into the SWLDA

Posted: 13 Feb 2020, 17:49
by sophie
Thank you so much Juergen! If I recorded using 8-channel cap and sampled at 256 Hz, then it would be 800-ms segments of data following each intensification from all 8 channels being fed into the SWLDA? and it will be downsampled to ~204 (256x0.8)sample?

Re: What being fed into the SWLDA

Posted: 14 Feb 2020, 07:34
by mellinger
The article mentions a sampling rate of 240Hz, and a decimation factor of 12. You have 204 samples in your epoch, sampled at 256Hz. The difference between 240 and 256 is roughly a factor of 12/13, so you should use a decimation factor of 13 rather than 12. 204/13 is roughly 15.7, so you should have 15 or 16 samples in the decimated epoch, depending on how your decimation algorithm behaves with regard to partial samples.

Don't forget to apply the moving average window before downsampling.

The article mentions a special set of electrodes. Whether you want to include all your electrodes depends on where they are located. You would not want to include frontal electrodes because you expect them to have more noise than signal.

Edit: remove erroneous reference to parietal electrodes

Re: What being fed into the SWLDA

Posted: 21 Feb 2020, 10:15
by sophie
Thank you, Juergen! One last question, you mentioned "would not want to include frontal/parietal electrodes because you expect them to have more noise than signal." Is that because the frontal site is near the eyes? Why would the parietal electrodes have more noise than signal? Why would we exclude it when that's where the P300 is the most robust at.

Re: What being fed into the SWLDA

Posted: 24 Feb 2020, 10:12
by mellinger
Sorry, I accidentally misrepresented the "p" in "Fp1"/"Fp2". Of course you need the parietal electrodes.
And yes, in frontal electrodes you have noise from eyes and muscles, that's why I suggested to exclude them.

I'm editing the above post to make sure nobody is misguided by my mistake.

-Juergen