JMIR Publications

ABSTRACT

Background:

Navigation remains a major challenge for people with visual impairment (PVI) as many assistive technologies (AT) fail due to limited contextual relevance and insufficient involvement of end-users during development. Traditional AT evaluation tools often miss the real-time, situational needs experienced during dynamic, everyday tasks.

Objective:

This study evaluated a mixed-methods framework combining task-based walking interviews with environmental scene metrics to capture real-time, context-dependent navigation needs of PVI. Findings were used to inform the design of our indoor navigation prototype, Edge A-Eye.

Methods:

Thirteen adults with visual impairments completed eight sequential indoor navigation tasks during a simulated optometry visit. Participants wore Pupil Labs Neon glasses while performing tasks and provided immediate verbal feedback after completing each task. Luminance, contrast, edge density and spatial entropy were extracted from video recordings to objectively characterize the scene. Multiple-choice responses were analyzed using descriptive statistics and correlations, while open-ended responses underwent thematic analysis.

Results:

Navigation needs varied substantially by context. Participants preferred early notifications (6–8 m) for entrance doors, but on-command cues for stairs, elevators, and exits. Across tasks, users consistently requested concise, actionable information such as door type, handrail position, or chair location rather than continuous descriptive narration. While human assistance remained preferred in complex areas, many participants expressed willingness to use AI-based guidance if reliability could be ensured. Descriptive comparisons across tasks indicated that higher environmental luminance and stronger object contrast were observed in scenarios where participants with low vision demonstrated higher navigation success.

Conclusions:

Task-based walking interviews combined with scene analysis provide ecologically valid insights into navigation behavior and user preferences. This methodology supports the development of adaptive, personalized assistive technologies and offers a scalable framework for future AT evaluation.