191 / 2019-12-30 03:40:00

Assistive Robot Multiple Action Movement Control Scheme Based on Binary Motor Imagery EEG

electroencephalogram (EEG) signal; motor imagery (MI); brain-computer interface (BCI); KUKA robotic arm; coding strategy

Electroencephalogram (EEG) signal reflects the functional state of human brain and the electrical activity of brain tissue. When people carry out motor imagery (MI) tasks, the rhythm responsible for motor perception in the brain will change accordingly. Through the brain-computer interface (BCI), the patterns can be recognized to help patients (stroke, paralysis, etc.) to undertake some grasping tasks and reconstruct their neural circuits by repeated exercises. In this paper, MI-EEG signal was used to control a grabbing module of a KUKA robotic arm and a soft touch manipulator. The subjects control the grasping module's movement in four directions by MI-EEG signal, and grab a bottle at target point. In offline experiment, the subjects' accuracy based on four-category MI brain activity was calculated and the four-classification model was constructed hereby. Then two types of the four-category with high accuracy called the subjects' optimal combination were determined to construct a two-classification model when undertaking the binary MI activity. The results of the online experiment show that both models realized the grabbing tasks. Compared with imagining four movements (with the classification accuracy of 38.8% and the total time of 44 steps), subjects achieved higher accuracy (with the classification accuracy of 84.2% and the total time of 41 steps) in less time by optimal binary MI combination. This paper poses a binary movement MI coding strategy to achieve four directions of the grabbing module, of which the performance is better than that of directly realizing the four directions with four-movement MI. Various functions can be achieved by fewer movements of MI, without increasing the difficulty of MI to achieve more complex applications of BCI. Repetitive exercises of MI can reconstruct patients' neural circuits, and the coding method can be used to further restore the particular MI of patients to accelerate rehabilitation in the future.