[Eeglablist] A question about the averaging for the P300 Speller

[Arnaud Delorme]

> Sorry for my bad English!
> A question about the averaging for a P300 work.
> The P300 Speller has  a 6 x 6 matrix of characters. Each row and each column are intensified by the system; the intensifications are presented in a random sequence.
> The user focuses attention on one of the 36 cells of the matrix, with the row and the column containing the character to be communicated constituting the rare set (target) , and the other 10 intensifications constituting the frequent set (no-target).
> In a first phase we go to train the classificator: in this phase the system knows if every epoch is  target or  no-target.
> In the online phase, when the user tries to select a character, the system doesn’t know if an epoch is target or no-target.
> In this situation, the  online phase, is useful to average epochs?

You can average epochs or probably better classify each epochs and then use a voting mechanism. You might want to check in the literature what makes the most sense.

Best wishes,

Arno