Channel selection procedures are essential to reduce the
curse of dimensionality in Brain-Computer Interface
systems. However, these selection is not trivial, due to
the fact that there are 2Nc possible subsets for an Nc
channel cap. The aim of this study is to propose a novel
multi-objective hybrid algorithm to simultaneously: (i) reduce
the required number of channels and (ii) increase the
accuracy of the system. The method, which integrates
novel concepts based on dedicated searching and deterministic
initialization, returns a set of pareto-optimal
channel sets. Tested with 4 healthy subjects, the results
show that the proposed algorithm is able to reach higher
accuracies (97.00%) than the classic MOPSO (96.60%),
the common 8-channel set (95.25%) and the full set of 16
channels (96.00%). Moreover, these accuracies have been
obtained using less number of channels, making the
proposed method suitable for its application in BCI
systems.
