Brain-Computer Interface (BCI)-Enhanced Control of Wearable Knee Exoskeletons for Walking Turning Assistance

Brain-Computer Interfaces (BCIs) offer promising solutions for mobility assistance for individuals with motor impairments. This paper presents a novel BCI-enhanced shared control approach for wearable knee exoskeletons for walk turning assistance. We introduce a neural architecture for real-time kinematic and intention detection design that leverages the electroencephalography (EEG) and inertial measurement unit (IMU) sensor fusion. The machine learning-based fusion system gradually reduces reliance on IMU data during training, enabling robust EEG-based control for real-time exoskeleton-assisted walking turning assistance. Extensive experiments demonstrate significant improvements in turning intention prediction and muscle activation reduction for exoskeleton control. The proposed design achieves a 91% accuracy in turning intention detection, a root mean square error of 13.7 deg in turning angle prediction, and reduces maximum muscle activation by approximately 10% in major muscles compared to traditional IMU-based exoskeleton control method.