Cortical Activity as a Marker for User Preference in Exoskeleton Personalization

Lower-limb exoskeleton performance greatly depends on personalization due to individual biomechanical and neurophysiological differences. Current methods struggle to effectively integrate user-specific needs into exoskeleton control. To address this, our proof-of-concept study explores EEG-based passive brain-computer interfaces (BCIs) for optimizing lower-limb exoskeleton settings. Fourteen participants wore a knee exoskeleton with adjustable pneumatic stiffness while performing a sinusoidal knee-tracking task. EEG, EMG, and knee kinematics were simultaneously recorded. Subjective difficulty ratings increased with higher stiffness levels, correlating closely with cortical activity changes, especially within alpha, beta, and gamma frequency bands. Classifiers trained on these neural patterns achieved, on average, 79 accuracy in distinguishing stiffness settings, demonstrating the potential of neuroadaptive strategies to personalize exoskeleton assistance dynamically based on brain responses.