Designing and Evaluating a Customizable Head-mounted Vision Enhancement System for People with Low Vision
Yuhang Zhao, Sarit Szpiro, Lei Shi, Shiri Azenkot · 2019 · ACM Transactions on Accessible Computing (TACCESS) · doi:10.1145/3361866
Summary
This paper presents ForeSee, a customizable head-mounted display (HMD) system that enhances the visual experience of people with low vision using real-time image processing techniques. The system offers five enhancement methods — magnification, contrast enhancement, edge enhancement, black/white reversal, and text extraction — in two display modes: Full Display (entire field of view enhanced) and Window Display (enhancement in a movable rectangular region). The research proceeded through three studies. The first study with 20 low vision participants evaluated which vision enhancements were effective across four viewing tasks (near- and far-distance image recognition and reading). A critical finding was that no single enhancement worked universally: participants needed to combine and adjust multiple enhancements to optimize their visual experience, and preferences varied dramatically even among people with the same eye condition. This motivated the second study, which designed two input modalities for real-time customization: speech commands and novel smartwatch-based touch gestures. The smartwatch interaction used a linear menu navigated by swiping, double-tap for selection, and a circular sliding gesture around the watch bezel to adjust enhancement levels. The third study with 11 low vision participants evaluated both input modes on a far-distance letter reading task, finding both effective — speech commands were faster and rated higher for effectiveness, while smartwatch gestures scored higher for usability and offered advantages in public or noisy settings where speech is inappropriate.
Key findings
The most significant finding was the critical importance of customization. More than 70 percent of participants in each task used magnification, but they combined it with different enhancements and at different levels depending on the task and their individual vision. Participants with the same diagnosed condition (e.g., three people with retinitis pigmentosa) chose entirely different enhancement combinations, demonstrating that diagnosis alone is a poor predictor of optimal visual enhancement — subjective visual experience matters more than clinical measures. The Window Display mode was preferred for near-distance tasks (avoiding peripheral distortion from full-screen magnification), while Full Display was preferred for far-distance tasks. Participants strongly desired the ability to save preferred enhancement combinations as presets and reload them automatically. When using the smartwatch, an unexpected challenge emerged: participants unconsciously moved their heads toward their hands while performing touch gestures, causing the magnified view to shift and making it harder to track the target. Speech commands avoided this problem but required memorizing exact command phrases, and participants found rigid command syntax frustrating. Both interaction modes received high effectiveness (6.09-6.91 out of 7) and usability scores (6.18-6.73). All participants could read letters with ForeSee that they could not read with their own corrective lenses alone.
Relevance
This research demonstrates the significant potential of consumer HMD technology as a low vision aid and provides detailed design guidelines for future systems. The central lesson — that customization is essential and clinical diagnosis is insufficient for predicting individual needs — has broad implications for assistive technology design. Rather than designing for specific eye conditions (the medical model approach), developers should empower users to adjust tools to their own subjective experience. The dual-input approach (speech plus smartwatch) exemplifies good accessible design by providing modality choice based on context: speech for efficiency, touch for privacy and noisy environments. The discussion of video see-through (VST) versus optical see-through (OST) smartglasses identifies VST as particularly promising for low vision because it allows complete image manipulation. For organizations developing wearable assistive technology, the finding that low vision users experienced no VR sickness (likely due to reduced sensitivity to frame rate and resolution) removes a common concern about HMD-based solutions.
Tags: low vision · head-mounted display · vision enhancement · magnification · edge enhancement · contrast enhancement · smartwatch · speech commands · customization · wearable technology