Methods for asynchronous and non-lnvasive eeg-based brain-computer interfaces, towards intel-ligent brain-actuated wheelchairs

Resumen

This thesis is aimed at developing methods for asynchronous and non-invasive EEG-based brain computer interfaces to enhance the robustness of brain-actuated devices. Within European MAIA project framework, these methods have been developed to enhance the robustness of a brain-actuated wheelchair. Three are the main contributions. First, the use of mental tasks transitions detection as inferred a priori information to guide postprocessing algorithms which goal is to denoise the decision making of the brain-computer interface system. Second, the use of a new feature extractor method for multi-class brain-computer interfaces with canonical solution that provides a reduced number of canonical discriminant spatial patterns and rank the channels sorted by power discriminability between classes. Third, the introduction of frames approach recognizing intermittent induced electroence-phalographic spatial patterns of amplitude modulation to guide a novel decision making process. The thesis also describes MAIA brain-actuated wheelchair architecture and the experiments carried out to assess its robustness. The thesis is accompanied with three demonstrations that allow to envisage the potential applications of this technology in real-life situations.