Identifying Motor Imagery-Related Electroencephalogram Features During Motor Execution

Abstract

Brain–computer interface technology facilitates communication and control of computers with brain signals. This technique uses motor imagery to enable a robotic arm to function as a third arm for the subject. During the process, the robotic arm must move in synchrony with the two human arms, and consequently motor imagery and motor execution must be performed simultaneously. In this study, we examined whether information related to motor imagery could be detected with an electroencephalogram during simultaneous measurement of motor imagery and motor execution. Our experiment included five participants who performed motor execution, and motor execution with motor imagery. To identify motor imagery-related features, we initially extracted event-related spectrum perturbation (ERSP) data and performed a t-test to examine significant differences using averaged-trial ERSP data. Subsequently, the data were classified with Fisher’s linear discriminant as the single-trial classification. Results revealed significant differences between the two movement conditions and the motor imagery-related features for each subject. The single-trial classification analysis demonstrated slightly higher accuracy than the chance level classification, but the difference was not significant. These results suggest that information related to motor imagery could possibly be decoded during motor execution, however performance improvement at the single-trial level will be necessary in future studies.