Kullback-Leibler Penalized Sparse Discriminant Analysis for Event-Related Potential Classification

A brain computer interface (BCI) is a system that measures brain activity and converts it into an artificial output which is able to replace, restore or improve any normal output (neuromuscular or hormonal) used by a person to communicate and control his/her external or internal environment. Thus, BCI can significantly improve the quality of life of people with severe neuromuscular disabilities [35]. Communication between the brain of a person and the outside world can be appropriately established by means of a BCI system based on eventrelated potentials (ERPs), which are manifestations of neural activity as a consequence of certain infrequent or relevant stimuli. The main reason for using ERP-based BCI are: it is noninvasive, it requires minimal user training and it is quite robust (in the sense that it can be use by more than 90 % of people) [34]. One of the main components of such ERPs is the P300 wave, which is a positive deflection occurring in the scalp-recorded EEG approximately 300 ms after the stimulus has been applied. The P300 wave is unconsciously generated and its latency and amplitude vary between different EEG records of the same person, and even more, between EEG records of different persons [18].