BLDA Approach for Classifying P300 Potential

Abstract

Brain-computer interface (BCI) provides the brain with a new communication and control channel for conveying messages and commands to the outside world. BCIs require effective on-line processing of EEG or P300 which appeared just 300ms after the event happened. In this paper, we use a method of Bayesian linear discriminant analysis (BLDA) to classify P300 signals from the data set of BCI Competition III. P300 are obtained when subjects were facing a screen on which flashed objects were displayed. Before training a classification parameter for the BLDA, several preprocessing operations were applied to the data including filtering, trial extraction et al. With the BLDA algorithm, the classification accuracy in our experiment could be up to above 80%, depends on the averaged times of trials and SNR of the data. In summary, it is suitable to use LDA in the P300-based BCI system, and LDA will be a promising method to extract feature for BCI designs.