Some questions about P300 Classifier

Forum for software developers to discuss BCI2000 software development
Post Reply
GiuseppeG
Posts: 1
Joined: 11 Sep 2015, 09:34

Some questions about P300 Classifier

Post by GiuseppeG » 23 Oct 2015, 07:27

Hello all.
I'm studying the source code of p300 classifier for my thesis, but I don't understand how p300 classifier uses SWLDA... Could someone help me?
Thank you very much!

tmokbel1965
Posts: 1
Joined: 23 Oct 2015, 17:09

Re: Some questions about P300 Classifier

Post by tmokbel1965 » 23 Oct 2015, 17:11

I am in the same boat. Kinda blown away by the entire complexity of it all. Can anyone explain?

pbrunner
Posts: 344
Joined: 17 Sep 2010, 12:43

Re: Some questions about P300 Classifier

Post by pbrunner » 07 Dec 2015, 14:28

The P300 Classifier is pretty simple. In a nutshell it performs a linear regression between the target and the non-target ERPs. As the discriminative information is only located a certain time periods, it applies a feature selection on the ERPs, hence the use of the stepwisefit function to determine the time period that are informative. This is implemented in BCI2000 in two steps:

1) Using the P300Classifier.exe to determine this linear regression.
From the copy spelling data, target and non-target ERPs are extracted and decimated in their temporal resolution to e.g., 20ms. A stepwisefit function selects the time period from each channel (up to a maximum of e.g. 60) that are informative. This yields a linear regression.

2) Using the output of the P300Classifier linear classifier in BCI2000 .
The linear regression is applied to the averaged ERPs associated to each intensified row and column. The row and column with the highest regression value is selected as the intended row and column, with the intersection defining the selected letter.

Post Reply

Who is online

Users browsing this forum: No registered users and 3 guests