A study on mental fatigue and workload electrophysiological markers : contributions to passive brain-computer interfaces ; Étude de corrélats électrophysiologiques pour la discrimination d'états de fatigue et de charge mentale : apports pour les interfaces cerveau-machine passives

Mental state estimation on the basis of cerebral activity and its resulting physiological activities has become a challenge for passive Brain-Computer Interfaces (BCI), in particular to address a need in neuroergonomics. This thesis work focuses on mental fatigue and workload estimation. Its purpose is to provide efficient and realistic processing chains. Thus, one issue was the modulation of workload markers as well as classification performance robustness depending on time-on-task (TOT). The impact of workload and TOT on attentional state markers was also assessed. For those purposes, an experimental protocol was implemented to collect the electroencephalographic (EEG), cardiac (ECG) and ocular (EOG) signals from healthy volunteers as they performed for a prolonged period of time a task that mixes working memory load and selective attention. Efficient signal processing chains that include spatial filtering and classification steps were designed in order to better estimate these mental states. The relevance of several electrophysiological markers was compared, among which spontaneous EEG activity and event-related potentials (ERPs), as well as various preprocessing steps such as spatial filtering methods for ERPs. Interaction effects between mental states were brought to light. In particular, TOT negatively impacted mental workload estimation when using power features. However, the chain based on ERPs was robust to this effect. A comparison of the type of stimuli that can be used to elicit the ERPs revealed that task-independent probes still allow very high performance, which shows their relevance for real-life implementation. Lastly, ongoing work that aims at assessing task-robust workload markers, as well as the usefulness of auditory ERPs in a single-stimulus paradigm will be presented as prospects. ; L'estimation de l'état mental d'un individu sur la base de son activité cérébrale et de ses activités physiologiques résultantes est devenue l'un des challenges des interfaces cerveau-machine (ICM) dites ...