Continuous sensorimotor rhythm based brain computer interface learning in a large population

Brain computer interfaces (BCIs) are valuable tools that expand the nature of communication through bypassing traditional neuromuscular pathways. The non-invasive, intuitive, and continuous nature of sensorimotor rhythm (SMR) based BCIs enables individuals to control computers, robotic arms, wheel-chairs, and even drones by decoding motor imagination from electroencephalography (EEG). Large and uniform datasets are needed to design, evaluate, and improve the BCI algorithms. In this work, we release a large and longitudinal dataset collected during a study that examined how individuals learn to control SMR-BCIs. The dataset contains over 600 hours of EEG recordings collected during online and continuous BCI control from 62 healthy adults, (mostly) right hand dominant participants, across (up to) 11 training sessions per participant. The data record consists of 598 recording sessions, and over 250,000 trials of 4 different motor-imagery-based BCI tasks. The current dataset presents one of the largest and most complex SMR-BCI datasets publicly available to date and should be useful for the development of improved algorithms for BCI control.