SSVEP-based active control of an upper limb exoskeleton using a low-cost brain–computer interface

摘要

Purpose For the robot-assisted upper limb rehabilitation training process of the elderly with damaged neuromuscular channels and hemiplegic patients, bioelectric signals are added to transform the traditional passive training mode into the active training mode. Design/methodology/approach This paper mainly builds a steady-state visual stimulation interface, an electroencephalography (EEG) signal processing platform and an exoskeleton robot verification platform. The target flashing stimulation blocks provide visual stimulation at the specified position according to the specified frequency and stimulate EEG signals of different frequency bands. The EEG signal-processing platform constructed in this paper removes the noise by using Butterworth band-pass filtering and common average reference filtering on the obtained signals. Further, the features are extracted to identify the volunteer’s active movement intention through the canonical correlation analysis (CCA) method. The classification results are transmitted to the upper limb exoskeleton robot control system, combined with the position and posture of the exoskeleton robot to control the joint motion of robot. Findings Through a large number of experimental studies, the average accuracy of offline recognition of motion intention recognition can reach 86.1%. The control strategy with a three-instruction judgment method reduces the average execution error rate of the entire control system to 6.75%. Online experiments verify the feasibility of the steady-state visual evoked potentials (SSVEP)-based rehabilitation system. Originality/value An EEG signal analysis method based on SSVEP is integrated into the control of an upper limb exoskeleton robot, transforming the traditional passive training mode into the active training mode. The device used to record EEG is of very low cost, which has the potential to promote the rehabilitation system for further widely applications.