A comparison of classification techniques for a gaze-independent P300-based brain-computer interface

This off-line study aims to assess the performance of five classifiers commonly used in the brain-computer interface (BCI) community, when applied to a gaze-independent P300-based BCI. In particular, we compared the results of four linear classifiers and one nonlinear: Fisher's linear discriminant analysis (LDA), stepwise linear discriminant analysis (SWLDA), Bayesian linear discriminant analysis (BLDA), linear support vector machine (LSVM) and Gaussian supported vector machine (GSVM). Moreover, different values for the decimation of the training dataset were tested. The results were evaluated both in terms of accuracy and written symbol rate with the data of 19 healthy subjects. No significant differences among the considered classifiers were found. The optimal decimation factor spanned a range from 3 to 24 (12 to 94 ms long bins). Nevertheless, performance on individually optimized classification parameters is not significantly different from a classification with general parameters (i.e. using an LDA classifier, about 48 ms long bins).