HaWCoS: The "Hands-free" Wheelchair Control System
Torsten Felzer, Bernd Freisleben · 2002 · Proceedings of the Fifth International ACM Conference on Assistive Technologies (Assets 02) · doi:10.1145/638249.638273
Summary
This paper introduces HaWCoS (Hands-free Wheelchair Control System), a system that enables people with severe physical disabilities to control an electrically powered wheelchair using muscle contractions from any single muscle group in their body, without requiring the use of their hands. The system works by monitoring the continuous stream of electromyography (EMG) signals from a chosen muscle via a surface electrode. These raw signals are processed to detect discrete contraction events — specifically "single clicks" (brief contractions) and "double clicks" (two rapid contractions). These events drive a finite state machine (FSM) that translates the binary input into four wheelchair commands: straight, left turn, right turn, and halt. The system has two operating modes: a simulation mode where the user can practice controlling a virtual object on screen, and a wheelchair mode where commands are sent to the actual wheelchair electronics via a PC-to-wheelchair interface. An adjustment phase at startup allows calibration of the electrode position and signal thresholds, typically taking less than two minutes. The prototype was implemented on an Invacare Grand GX34K-Pro wheelchair equipped with Dynamic Controls DX module, with the computer recording EMG amplitude at 100 Hz and updating wheelchair commands every 0.1 seconds. The research was motivated by the reality that many wheelchair users cannot reliably use the standard joystick control, particularly those with severe motor impairments affecting hand function.
Key findings
Practical experiments demonstrated that HaWCoS is a viable alternative to traditional joystick control, with an overhead of approximately 50% or less in completion time compared to joystick steering. Three route types were tested: a long outdoor distance (312 seconds by joystick vs. 441 seconds by HaWCoS, 41% overhead), a medium indoor distance (67 seconds vs. 98 seconds, 46% overhead), and a short "shunting" route requiring precise maneuvering through doorways (under 20 seconds vs. under 30 seconds, 50% overhead). Notably, the variation in HaWCoS timing was small (under ±5% after practice) for the medium and short routes, though the long outdoor route showed greater variability. The system proved extremely easy to learn — users needed only 15-20 minutes to become proficient with the EMG-based control. The authors note that while HaWCoS only allows movement in "STRAIGHT" or "TURN" states (not both simultaneously), this constraint actually simplifies operation and reduces errors. A key practical finding was that the system requires reliable control over only one muscle group, making it suitable for people with very severe physical disabilities who retain voluntary control of even a single muscle, such as a facial muscle or an eyebrow.
Relevance
This paper demonstrates an innovative approach to wheelchair accessibility that remains relevant today as EMG-based and brain-computer interface technologies continue to advance. The fundamental design insight — reducing complex continuous control to simple binary inputs (single and double clicks from any muscle) — is an elegant solution for users with the most severe motor impairments. For accessibility practitioners, the key takeaway is that effective assistive technology does not always require complex input mechanisms; a well-designed finite state machine can translate minimal input into meaningful control. The 50% time overhead compared to joystick control is a reasonable trade-off for users who have no alternative means of wheelchair operation. The system's limitation is that it addresses only wheelchair navigation and not computer interaction, though the authors note the same EMG input technique could be extended to character selection and general computer use. The paper also highlights important practical considerations for EMG-based systems, including electrode placement calibration and artifact rejection from involuntary movements like convulsions.
Tags: electromyography · wheelchair control · physical disability · alternative input · assistive technology · hands-free interaction · EMG signals