Use of the Choquet integral for combination of classifiers in P300 based brain-computer interface

One of the key issues in the development of brain-computer interfaces (BCIs) is the improvement of their current information transfer rate. In order to achieve that objective at least two aspects of BCI design should be considered: classification accuracy and protocol specification. In this paper we show how combination of classifiers using fuzzy measures and the Choquet integral can be applied to the context of visual P300 BCI in order to lower the number of misclassifications. Results of an offline analysis are provided and possible benefits in terms of the information transfer rate are briefly discussed.