Wavelet-Based Semblance Methods to Enhance the Single-Trial Detection of Event-Related Potentials for a BCI Spelling System

Algorithm 2

Semblance-based ERP window selection by channels (SEWS-1).

(1)	Input: given the EEG signal matrix X, with C channels and T temporal samples
(2)	Output: the bounds of the temporal window and for each C channel
(3)	Set the window threshold ,
(4)	for to C do
(5)	Compute the average of responses belonging to the target class for channel c
(6)	Compute the average of responses belonging to the nontarget class for channel c
(7)	Compute the and using equation (2)
(8)	Compute the semblance using equation (4)
(9)	Compute the dot product D using equation (5)
(10)	Compute the standard deviation of D over the scales and standardise it between 0 and 1
(11)	The limit is given by the first t from the left which meets
(12)	The limit is given by the first t from the right which meets
(13)	end for