Channel Selection and Dimensionality Reduction using Genetic Algorithms for a P300 Brain Computer Interface

M. PACHECO; Y.V. ATUM; J. A. BIURRUN MANRESA; C.B. TABERNIG; R. ACEVEDO

Córdoba

Congreso; XXI Congreso Argentino de Bioingeniería y X Jornadas de Ingeniería Clínica; 2017

Sociedad Argentina de Bioineniería

A Brain Computer Interface (BCI) is a device that provides a direct communication pathway between brain and computer. In this work, a P300-based BCI was implemented, in which feature selection stage consisted on channel selection and dimensionality reduction using Genetic Algorithms (GA).The classification stage was implemented using Fisher?s Linear Discrimination Analyses (LDA). A dataset of input patterns was generated from a database of EEG recordings from 18 healthy volunteers in order to train and test the proposed configuration. The addition of the GA as a channel selection resulted in a significant improvement of classification performance and in a significant reduction in the number of features. Results showed that four of the ten channels are the ones that really contribute to the classification of patterns with and without P300. In this case, a (77,9 ±7,6)% of accuracy was obtained with patterns of 256 characteristics in comparison with the application of selection of characteristics with GA for these four channels, in which case the accuracy was (84,9 ±6,5)% and the average number of channels (125,3 ±14,8).