Towards Usable Electroencephalography-based Brain-Computer Interfaces

Brain-Computer Interfaces (BCIs) are systems that can translate brain activity patterns of a user into messages or commands for an interactive application. Such brain activity is typically measured using Electroencephalography (EEG), before being processed and classied by the system. EEG-based BCIs have proven promising for a wide range of applications ranging from communication and control for motor impaired users, to gaming targeted at the general public, real-time mental state monitoring and stroke rehabilitation, to name a few. Despite this promising potential, BCIs are still scarcely used outside laboratories for practical applications. The main reason preventing EEG-based BCIs from being widely used is arguably their poor usability, which is notably due to their low robustness and reliability, as well as their long calibration and training times. The research presented in this manuscript aims at addressing these different points in order to make EEG-based BCIs usable, i.e., to increase their efficacy and efficiency. In particular, we present a set of contributions towards this goal 1) at the EEG signal processing and classification level, to robustly decode EEG signals and translate them into commands, 2) at the user training level, to ensure that users can learn to control a BCI efficiently and effectively, and 3) at the usage level, to explore novel applications of BCIs for which the current reliability can already be useful. First, in terms of EEG signal processing tools, we proposed a number of methods to improve BCI reliability, despite EEG signal variability, poor signal-to-noise ratio and high sensitivity to artifacts, as well as to reduce BCIs calibration times. More precisely, we complemented traditionally used features by exploring alternative representations of EEG signals. We also explored and designed regularized spatial filters to learn more robust and stable features. Finally, we propose algorithms to reduce BCIs calibration times, i.e., to calibrate BCIs with as few examples of EEG signals from the target user as possible, by re-using data from previous users or by generating artificial EEG signals. Altogether, these methods enabled an increased BCI classification accuracy, i.e., efficacy, and BCI calibration with much less data than standard approaches do, thus improving their efficiency. Second, rather than improving EEG signal processing alone, we advocate that BCIs can also be made more usable by guiding users to efficiently learn BCI control mastery. Indeed, BCI control is known to be a skill that needs to be learnt. A study of models and guidelines from educational sciences enabled us to identify many theoretical limitations of current standard BCI training approaches, thus highlighting the need for alternative ones. In particular, educational sciences recommend to train people with adapted and adaptive training tasks, using explanatory feedback in motivating environments. In contrast standard BCI training protocols are commonly fixed, repetitive, rather boring and provide purely corrective feedback. To address these limitations, we studied what kind of users manage to use a BCI and why. We also explored new feedback types, in particular richer feedback, multi-user feedback and tactile feedback to help users to learn BCI control skills more efficiently. Overall, our studies identified some cognitive (notably spatial abilities) and personality factors playing a major role in mental imagery-based BCI performances. They also revealed that both tactile feedback and social presence can improve BCI efficacy. Finally, BCIs can be made more usable by being used for other applications than communication and control. To this end, we notably explored the use of BCIs for neuroergonomics, i.e., using brain signals to passively estimate some of the relevant user's mental states during human-computer interaction, in order to assess the ergonomic qualities of this interface. In particular, we showed that one can estimate mental workload during complex 3D manipulation and navigation tasks in order to assess or compare interaction techniques and devices. We have also been able to study stereoscopic displays by estimating visual comfort in EEG signals. Another usage of BCIs, that we found promising and useful, is real-time brain activity and mental state visualization. We designed a number of devices based on augmented reality and/or tangible interfaces to enable novice users to visualize their own brain activity or mental states in real-time, with potential applications in fields as wide and diverse as education, self-awareness or well-being. Overall this work contributed novel methods and approaches to make EEG-based BCIs more usable, as well as new knowledge that could be used to further improve them in the future. This manuscript also proposes some perspectives and directions that could be worth exploring to that end. BCIs show a huge potential for research and applications, and we hope our research will contribute to turn this potential into realities.