Predicting emg envelopes of grasping movements from eeg recordings using unscented kalman filtering

Electromyographic (EMG) control of prosthetics is well established both in research and cl inical settings. However, it remains unclear how much of the EMG information can be predicted from the electroencephalographic (EEG) signals, and used instead, for control. In this study, we used a dataset that contains simultaneously acquired EEG and EMG signals of 31 subjects performing 33 grasping conditions, and applied unscented Kalman filtering (UKF) to continuously predict the EMG grasping envelopes from the low-frequency (0.1-2 Hz) EEG. We achieved higher prediction accuracy for intermediate grasps compared to power or precision grasps. Our findings indicate the feasibility of continuously predicting EMG envelopes of grasping movements from EEG signals. Keywords EEG, EMG, grasping, UKF