Multilinear Discriminative Spatial Patterns for Movement-Related Cortical Potential Based on EEG Classification with Tensor Representation

Algorithm 1

Training procedure of multilinear discriminative spatial patterns.

	Input: a training set of EEG signals, with the class label .
	Output: the optimal projection matrices
(1)	initialize (I means identity matrix).
(2)	for t = 1, 2, ..., do
(3)	fordo
(4)	.
(5)	.
(6)	.
(7)	.
(8)	.
(9)	.
(10)	optimize through .
(11)	if t > 2 and then
(12)	.
(13)	break
(14)	end if
(15)	end for
(16)	end for