8-channel EMG, EEG upper limb gesture data.
This dataset includes EMG and EEG data acquired using the Myo armband and OpenBCI Ultracortex IV. The dataset is intended for the intuitive control of a rehabilitation device.
Steps to reproduce
Eleven able-bodied subjects (seven males, right-handed, one male, left-handed, and three females, right-handed; ages 25 ± 5.5 years) without known neuromuscular dysfunction participated in the data acquisition. Thalmic Labs' array-based 8-channel Myo armband with a sampling frequency of 200Hz was used to acquire the EMG from the proximal part of the subject's forearm. The OpenBCI Ultracortex "Mark IV" was used to acquire the EEG data using the 8-channel option without a daisy connection. The device has a sampling frequency of 250Hz and is placed on F3, F4, C3, CZ, C4, P3, PZ and P4 with the earlobes as reference. EMG and EEG data were acquired simultaneously on independent computers using a custom design software with the same acquisition protocol. The offline data collection per trial is started manually with the target gesture displayed across the screen (24-inch) 1.5m from the subject, with a simultaneous acoustic tone (1kHz, 20ms) to signify the start of the hold while performing motor imagining (MI) task after the visual cue. Subjects were instructed to perform imaginary and grasp tasks during the hold period. The gesture is held for 5 seconds, followed by a second sound indicating a stop hold grasp and the beginning of a 3-second rest interval. A total of six repetitions were performed per gesture trial, totaling 42 trials for a single subject. The last two repetitions involved the subject standing up with the forearm parallel to the ground to mimic a real-life scenario. The manual start and stop sequence followed in the acquisition of the data set ensured that the dataset was balanced in all classes.
National Research Foundation of Korea