Computational Intelligence and Neuroscience vol:2011 pages:1-12
In this paper, we report on tests with the P300 Brain-Computer Interface (BCI) typing paradigm on neurological patients suffering from motor and speech disabilities. We investigate the accuracy of different classifiers: Fisher's Linear Discriminant Analysis (LDA), Bayesian Linear Discriminant Analysis (BLDA), Stepwise Linear Discriminant Analysis (SLDA), a method based on Feature Extraction (FE), linear Support Vector Machine (SVM), Gaussian kernel Support Vector Machine (nSVM) and multi-layer perceptron (NN). Tests were performed on patients suffering from a Amyotrophic Lateral Sclerosis (ALS), a middle cerebral artery (MCA) stroke, and a Subarachnoid Hemorrhage (SAH), both in on-line and in off-line mode. Our results show that BLDA, in general, yields a higher classification accuracy than the other classifiers, except for nSVM in some cases, at least for our group of patients.