Classification And Recognition Of P300 Evoked Potentials And Its Application In BCI

A brain-computer interface (BCI) is a real-time communication system in which messages or commands sent by the user do not pass through the brain's natural output pathways. The brain waves which could be used in BCIs include EEG, EcoG et al. There are kinds of BCIs based on the different brain signals. People have designed BCIs based on the user's response to specific sensory events or Event Related Potential such as P300. P300 appeared just 300ms after the event happen and could be obtained when subjects were facing a screen on which flashed subject were displayed. In this study, we will discuss different kinds of electrode configurations and machine learning algorithms to find out the best way to classify the P300 signals.In this thesis, we study methods of FLDA( Fisher's linear discriminant analysis) , BLDA ( Bayesian linear discriminant analysis) and SVM (Support vector machine) which are used to classify P300 signals. Before training a classification parameter for the algorithm, several preprocessing operations were applied to the data including filtering, downsampling, single trial extraction, electrode selection et al. Experiments indicate that LDA has the advantages of easy implementation and low computational cost, and is suited for BCI based on P300. The classification accuracy depends on the configuration and number of electrodes, the SNR of the data and the algorithm we selected. Further work will include the algorithm improvement to yield higher classification accuracy.