Event-related desynchronization (ERD) may not be correlated with motor imagery BCI performance

Abstract

Motor imagery brain-computer interfaces (BCIs) can control external machines with neurophysiological signals during limb movement imagination without real movements. An electroencephalography (EEG)-based motor imagery task has notable characteristics of event-related (de) synchronization (ERD/ERS) in specific frequency bands, including alpha and beta rhythms in the sensorimotor area. Based on this phenomenon, motor imagery features are extracted typically using common spatial patterns (CSPs). However, some researchers have reported that ERD features have severe inter-subject variation. In this study, we investigated the correlation between various ERD features and classification accuracy during a motor imagery task. We found that ERDs may not be useful in estimating classification accuracy, although they are representative features of a motor imagery task.