M.: Pr. Pascal GUITTON President M.: Pr. Bruno ARNALDI Directeur de thèse

A Brain-Computer Interface (BCI) is a communication system which enables its users to sendcommands to a computer by using brain activity only, this brain activity being measured, gen-erally by ElectroEncephaloGraphy (EEG), and processed by the system.In the first part of this thesis, dedicated to EEG signal processing and classification tech-niques, we aimed at designing interpretable and more efficient BCI. To this end, we first pro-posed FuRIA, a feature extraction algorithm based on inverse solutions. This algorithm canautomatically identify relevant brain regions and frequency bands for classifying mental states.We also proposed and studied the use of Fuzzy Inference Systems (FIS) for classification. Ourevaluations showed that FuRIA and FIS could reach state-of-the-art results in terms of classi-fication performances. Moreover, we proposed an algorithm that uses both of them in orderto design a fully interpretable BCI system. Finally, we proposed to consider self-paced BCIdesign as a pattern rejection problem. Our study introduced novel techniques and identified themost appropriate classifiers and rejection techniques for self-paced BCI design.In the second part of this thesis, we focused on designing virtual reality (VR) applicationscontrolled by a BCI. First, we studied the performances and preferences of participants whointeracted with an entertaining VR application, thanks to a self-paced BCI. Our results stressedthe need to use subject-specific BCI as well as the importance of the visual feedback. Then,we developed a VR application which enables a user to explore a virtual museum by usingthoughts only. In order to do so, we designed a self-paced BCI and proposed an interactiontechnique which enables the user to send high-level commands. Our first evaluation suggestedthat a user could explore the museum faster with this interaction technique than with currenttechniques.tel-00356346,version2-29Jan2009