Jorge Del Río Vera BCI competition 2003
data set iv (berlin)
Algorithm
The algorithm used is based in a MLP neural network and a key
feature extraction based in PCA (principal component analysis).
The principal component analysis is computed for each channel, then
the second principal mode is used (I have not used the first because
it contains the normal activity of the brain instead the features to
discriminate between left or rigth hand). This mode is correlated
with the signals (only the signals of the left side of the brain are
used, beacuse the information of the laterallity is contained in the
second principal mode) and the delay of the maximun is used for
training the net (a 12 component vector). The net used is a MLP
with two inner layers in order to accelerate the convergence (10 and
5 neurons respectively) and the output layer has two neurons.
Results
Results are given in a matlab data file, it have the same order than
the test signals and the class is given by a 0 or a 1.