Single-trial discrimination of type and speed of wrist movements from EEG recordings

Abstract

The study explored the possibility of identifying movement type and speed from EEG recordings. EEG signals were acquired from 9 healthy volunteers during imagination of four tasks of the right wrist that involved two speeds (fast and slow) and two types of movement (wrist extension and rotation), each repeated 60 times in random order. Average movement-related cortical potentials (MRCPs) were compared among the four tasks. Moreover, single-trial classification was performed using the rebound rate of MRCP and the power in the mu and beta bands as features. The rebound rate of the average MRCPs was greater for faster than for slower movements but did not depend on the type of movement. Accordingly, pairs of tasks executed at different speeds led to lower misclassification rate than pairs of tasks executed at the same speed. The average misclassification rate between task pairs was 21+/-2% for the best channel and task pair. The task parameter speed can be discriminated in single-trial EEG traces with greater accuracy than the type of movement when tasks are executed at the same joint. The speed of movement execution may be included among the variables that characterize imagined tasks for brain-computer interface applications.