NEARBY

Brain-computer interfaces, or BCIs for short, offer a promising possibility for human-machine interaction based on brain signals, especially as an interface for operating assistance systems for physically impaired persons, but also generally for controlling technical systems without the use of hands. Despite the promising benefits of this technology, it is rarely used outside of controlled laboratory conditions. This is mainly due to the lack of reliability of the systems. The measured brain activity not only differs greatly between people but also varies within the same person, depending on the mental or physical state at the time of use. This variability between and within the brain activity of BCI users represents one of the greatest challenges in its application in everyday scenarios.

The NEARBY project aims to develop variability-free BCI systems for out-of-the laboratory usage. To this end, a comprehensive database will be created in which EEG data from various test subjects is recorded over longer periods of time under different conditions and in different environments. The aim is to better understand the variability of the data under different conditions and to develop new algorithms that can reduce or even completely suppress this variability. Machine learning methods shall be used for noise suppression, and robustness against changing conditions is to be improved through deep and meta-learning algorithms on the shared data structure.

Variability in brain activity is one of the biggest barriers on the path of BCIs from the lab to everyday use. The NEARBY project lays the foundation for understanding this variability by collecting an extensive database of EEG data with standardized recording protocols under a wide variety of conditions and, thereby, aims at providing new approaches for developing methods to reduce variability.

This shall enable the development of new, more robust BCI systems that are also suitable for non-medical purposes, e.g., for hands-free interaction in industrial scenarios or video games. Existing BCI approaches can also be extended by using variability reduction to design more robust interaction principles. The BCIs will be tested in two application scenarios from the fields of human-robot interaction and exoskeleton-based therapy.