Feature selection based on modified genetic algorithm for optimization of functional near-infrared spectroscopy (fNIRS) signals for BCI

One of the pivotal issues which must be tackled when an effective brain-computer interface (BCI) is to be designed, is to reduce the enormous space of features extracted from fNIRS signals. BCI researchers often use genetic algorithms (GA) as the technique to extract features. The classic genetic algorithm obtains a feature set with the high classification accuracy; however, it is unable to create a reliable classifier due to the large dimensionality of this set. Herein, we propose a refined version of the genetic algorithm, which chooses a small number of features and determines the optimal feature combination for classification of fNIRS signals for a BCI system.