Fatigue-adaptive EMG interface for real-time asynchronous wheelchair navigation

Individuals with spinal cord injuries (SCIs), neuromuscular disorders, or stroke-related impairments often face difficulties in operating powered wheelchairs with conventional control interfaces. This study aimed to develop and evaluate a lightweight surface electromyography (sEMG)-based system that enables intuitive and reliable wheelchair navigation using residual neck and shoulder muscle activity. The system employed three electrodes positioned over the right and left trapezius and sternocleidomastoid (SCM) muscles. Signals were processed in real-time using an ESP32 microcontroller, eliminating the need for external hardware. A single feature-standard deviation (SD) of the sEMG signal-was extracted from 3-second windows to detect commands. A dynamic thresholding mechanism was implemented to compensate for muscle fatigue without increasing computational demand. Five wheelchair navigation commands were classified. Experimental validation was conducted with 20 participants, including 10 participants with disabilities (PwDs) and 10 healthy controls. Healthy participants achieved 100% accuracy, with an average response time of 3.14 s and an information transfer rate (ITR) of 44.40 bits/min. PwDs achieved 96.75% accuracy, an average response time of 3.20 s, and an ITR of 38.59 bits/min. All participants reported the system to be safe, comfortable, and easy to use. The proposed sEMG-based wheelchair control system provides a practical, real-time, and cost-effective solution for individuals with limited upper- and lower-limb mobility. Its minimal setup, low computational complexity, and adaptive fatigue compensation make it suitable for daily use, offering an accessible alternative to conventional wheelchair control methods.