Abstract:
Mobility impairments caused by spinal cord injuries, stroke, and degenerative disorders limit independ
ence for millions worldwide. This study presents a low-cost lower-limb exoskeleton that decodes user intention in
real time using electroencephalographic (EEG) and electromyographic (EMG) signals. The system employs
a PIC16F877A microcontroller executing adaptive impedance control and gait-phase recognition, with a Kalman
filter combining inertial, torque, and bio-signal inputs to generate smooth hip–knee trajectories. Bench-top and
hardware-in-the-loop simulations classified four locomotor commands – stand, sit, walk, stop – with 96 %
accuracy and a maximum 85 ms latency. The lightweight 3.8 kg aluminium–carbon frame, powered by back
drivable BLDC motors, reproduced natural hip excursions of 15–35° at 1.2 s cadence. The results validate the
feasibility of embedding edge AI for assistive gait restoration in resource-constrained clinical contexts.
Keywords: exoskeleton, EEG, EMG, intent recognition, adaptive control, rehabilitation.
