Single Electrode Energy on Clinical Brain–Computer Interface Challenge

Abstract

Brain–computer interfaces (BCIs) can control and send messages to electronic devices by classifying brain signals. Several applications have been developed with this principle, and BCIs are constantly growing due to their broad applicability. One of the brain signals acquisition methods is the electroencephalogram (EEG). On EEG-BCI, the signal commonly passes through the stages of temporal and spatial filtering, feature extraction, and classification. The Clinical BCI Challenge 2020 occurred to encourage new techniques and assess the state-of-the-art for motor classification of hemiparetic stroke patients. This work presents how we won the competition and the new approach single electrode energy (SEE) developed for this competition. The model proposed here took first place overall in the competition with a score equal to 15.500, which is 7.5% better than the score reached by the second-place team. This result shows that SEE is the best technique to classify the clinical data from the BCI Challenge among the methods presented in this competition and indicates a new way to classify the movements imagined by hemiparetic patients with stroke.