← All reviews

Gaze Typing Through Foot-Operated Wearable Device

Vijay Rajanna · 2016 · ASSETS '16: Proceedings of the 18th International ACM SIGACCESS Conference on Computers and Accessibility · doi:10.1145/2982142.2982145

Summary

This paper presents a dwell-free gaze typing system that combines eye tracking with a foot-operated wearable device to enable text entry for people with motor impairments affecting arms or spine. Traditional gaze typing requires users to "dwell" (fixate) on virtual keyboard keys for 150-200ms to select them—a method that causes visual fatigue during extended use and results in lower typing speeds and higher error rates. The proposed system eliminates dwell time by separating gaze (pointing) from selection (activation). Users look at the desired key on a virtual keyboard while the eye tracker (Tobii EyeX) tracks their gaze position, then press a pressure pad attached to a foot-worn wearable device to confirm selection. The system comprises three components: a customized virtual keyboard based on the open-source OptiKey, a gaze interaction server that connects the components via Bluetooth, and a 3D-printed wearable housing containing a force-sensitive resistor and microcontroller. The virtual keyboard design underwent multiple iterations based on usability testing. Key refinements included making high-frequency letters larger, keeping the keyboard size optimized to minimize head tilting, moving infrequently used keys to a secondary screen, and making functional keys (space, enter, backspace) prominent. Visual feedback highlights the currently gazed key with a red border and confirms selection with a blue background.

Key findings

A preliminary evaluation with two wheelchair users with limited motor functions achieved a mean typing speed of 6.23 Words Per Minute—comparable to the 6.97 WPM typically achieved by non-disabled users with standard dwell-based gaze typing. Error metrics were near-ideal: Key Strokes Per Character was 1.07 (ideal 1.0), and Rate of Backspace Activation was 0.07 (ideal 0.0), indicating high accuracy. Formative testing with four non-disabled participants revealed important design insights: simply enlarging the entire keyboard does not improve performance and actually causes strain from constant head tilting. The optimal approach keeps overall keyboard size minimal while enlarging only high-frequency keys—a finding that informed the final design. The 3D-printed wearable housing enables customization for individual users and different interaction paradigms. While foot interaction suits users with arm or hand impairments, the same device paradigm could support hand operation for users with other types of impairments, making the system adaptable to diverse motor abilities.

Relevance

This research addresses a critical gap in assistive text entry by tackling the fatigue problem inherent in dwell-based gaze typing. For people who rely on gaze as their primary computer input—often due to severe motor impairments from spinal cord injury, ALS, or similar conditions—reducing fatigue directly impacts how long they can work and communicate each day. The multimodal approach (gaze + foot) demonstrates a valuable design pattern: separating pointing from activation allows each modality to do what it does best. This principle applies broadly to assistive technology design. The 3D-printed, customizable form factor also points toward a future where assistive devices can be rapidly personalized rather than one-size-fits-all. For practitioners, the keyboard design findings are immediately applicable: optimize key sizes based on frequency rather than uniformly enlarging interfaces, and minimize the need for head movement. The system builds on existing open-source tools (OptiKey, commodity eye trackers), suggesting that similar multimodal assistive solutions could be developed without extensive custom hardware.

Tags: gaze typing · eye tracking · wearable devices · foot interaction · motor impairment · text entry · assistive technology · alternative input · dwell-free interaction · 3D printing