See-Thru: Towards Minimally Obstructive Eye-Controlled Wheelchair Interfaces
Corten Clemente Singer, Björn Hartmann · 2019 · Proceedings of the 21st International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/3308561.3353802
Summary
This paper presents See-Thru, an eye-tracking interface for controlling power wheelchairs that replaces the conventional obstructive computer screen with a transparent wire-frame device containing spatially arranged LEDs. The fundamental problem with existing eye-controlled wheelchair systems is that they mount a computer screen directly in the user's field of view (FOV) to display gaze-activated control buttons and provide visual feedback. While this screen enables interaction, it blocks the user's view of the path ahead — a critical safety issue when navigating a wheelchair. Some systems overlay a live video feed of the forward view on the screen, but this creates a disorienting spatial mismatch between the video and the user's peripheral perception of the real environment. See-Thru addresses this by using an acrylic frame (approximately the size of the eye tracker's calibration area) with LEDs embedded along the perimeter while the center remains transparent. The LEDs serve dual purposes: they are gaze targets (the user looks at an LED to issue a command) and feedback indicators (LEDs change color to show system state). Green LEDs indicate available drive directions, red LEDs indicate the wheelchair will halt, and cyan LEDs indicate clutch controls for switching between Drive and Rest states. The system uses a Tobii Eye Tracker 4C (), communicates with a Windows Surface Pro tablet (hidden from view), and controls a Permobil C400 power wheelchair via standard RNET software. A finite state machine manages two main states: Rest (only one clutch LED active, preventing accidental movement) and Drive (four directional LEDs plus a clutch LED to return to Rest). A four-step clutching sequence requiring consecutive LED activations around the frame prevents unintentional state changes.
Key findings
Seven power wheelchair users with diverse disabilities (systemic mastocytosis, spinal muscular atrophy type 2, cerebral palsy, dwarfism, scoliosis, C4/C5 quadriplegia) evaluated See-Thru against a screen-based control in three navigation tasks: straight-line driving (21ft forward, 16ft backward within 5.5ft boundaries), curved path with 180-degree turn, and figure-eight around two obstacles. All participants completed all tasks with both interfaces, most having never used an eye tracker before. Task completion times were comparable between conditions (slightly faster with See-Thru in two of three tasks), but participants made fewer errors with See-Thru across all tasks — significantly fewer in Task 2 (0.43 vs 1.29 errors, p<.05). Four of seven participants preferred See-Thru, two had no preference, and one preferred the screen. Participants who preferred See-Thru cited the ability to view the environment naturally without the distraction of a screen, with one noting they felt "super focused on the screen rather than my surroundings" and another feeling "slightly claustrophobic" with the screen in front of their face. The spatial mismatch between the video feed and reality was specifically criticized. All participants found See-Thru easy and fun to learn, with each able to drive independently within minutes. Participants generally felt the system was best suited for private/home environments rather than crowded public spaces, though one participant with cerebral palsy expressed immediate interest in using it in their daily life.
Relevance
This research addresses a critical design tension in eye-controlled assistive technology: the eyes must simultaneously serve as both the input device (sending commands) and the sensory organ for environmental awareness (observing the path ahead). When a screen blocks the user's FOV, it forces an impossible dual-attention task — watching the interface and watching where you're going. For people with severe motor disabilities who rely entirely on eye control for mobility, this is not merely an inconvenience but a safety issue. The See-Thru approach of using peripheral LED indicators on a transparent frame is an elegant solution that preserves the user's direct view of the environment while still providing sufficient feedback. For accessibility practitioners and assistive technology designers, the key insight is that interface paradigms designed for desktop computing (screen-based gaze interaction) do not automatically transfer to navigation contexts. The finding that fewer errors occurred with See-Thru — despite comparable completion times — suggests that the unobstructed view provided better environmental awareness, supporting safer navigation. The study's inclusion of participants with progressive conditions (like spinal muscular atrophy) who will eventually need eye-gaze control as their motor abilities decline makes the findings directly relevant to future users of this technology.
Tags: eye tracking · eye-gaze control · power wheelchair · motor disability · assistive technology · field of view · navigation · user interface design · user study