Decoding hand movement velocities from EEG signals during a continuous drawing task

In brain-computer interface (BCI) studies, decoding neural activities representing limb movements is the key of motor prostheses controlling. So far, most of these works have been based on invasive approaches. But a few researchers tried to decode kinematic parameters of single hand from magnetoencephalogram (MEG) or electroencephalogram (EEG) signals during center-out reaching tasks. Yet whether and how EEG activities might be related to hand velocities during continuous drawing task is still unclear. Here we applied spatial filtering to multi-channel EEG in different frequency bands and then employed a Kalman smoother to decode hand movement velocities during a two-dimensional drawing task. The mean correlation coefficients between measured and decoded velocities ranged from 0.35~0.83 for the horizontal dimension and 0.11~0.45 for the vertical dimension. These results indicated that continuous neural control of motor prostheses may be realized by recoding EEG with visual motor task.