Brain computer interaction: information theoretic performance measures and protocol optimization. We have modeled synchronous BMI systems with discrete transducer outputs using an additive white Gaussian noise channel with features represented by pulse amplitude modulation. Using this model, we have been able to compare various information theoretic performance measures (e.g. Wolpaw, Nykopp), to compute the optimal protocol speed for average-trial protocols, and to determine the optimal number of mental tasks used in the protocol. For non- synchronous protocols, information theoretic performance measures are not well adapted and we are developing appropriate novel performance estimation methods. We are also working on the definition of information theoretic performance measures for synchronous BMI systems with continuous transducer outputs.

Past research topics:

• Research on direct & inverse problems ( R. Rytsar);
• Research on information-transfer rate (J. Kronegg).