Smartphone-Based Assistance for Blind People to Stand in Lines
Seita Kayukawa, Hironobu Takagi, João Guerreiro, Shigeo Morishima, Chieko Asakawa · 2020 · Extended Abstracts of the 2020 CHI Conference on Human Factors in Computing Systems (CHI EA '20, Late-Breaking Work) · doi:10.1145/3334480.3382954
Summary
This CHI 2020 Late-Breaking Work is the preliminary study that seeded the CHI 2021 LineChaser paper by the same research group. Where the fuller LineChaser system would later combine line-end finding with AR-marker-based localisation and a dual audio/vibration interface, this earlier paper takes a simpler cut: the blind user has already reached the end of a line, and the system only has to help them follow the person in front. An off-the-shelf iPhone 11 Pro uses its RGB camera plus infrared depth sensor to detect the nearest pedestrian with YOLOv3-tiny at 2–3 fps and estimate the distance from the target's centroid, reliable between 0.2 m and 6 m. Three vibration patterns — a long pulse for 'stop' (target within 50 cm), a two-pulse for 'move forward' (target beyond 50 cm), and a rapid tick for 'obstacle' (anything within 50 cm) — are the entire feedback interface; no audio, no text-to-speech, no map. The authors ran a user study with six totally blind cane users (P1–P6, ages 22–33), simulated a shopping-mall queue with four sighted 'extras' moving intermittently plus a 60 dB ambient-noise track, and asked participants to follow the line to a reception desk across two line-movement conditions (extras leaving one by one vs two at a time).
Key findings
All six participants successfully followed the line to the goal, stopping just behind the target on 84.4% of line-movement events (108/128) and within the acceptable position zone on 94.5% (121/128) when slight lateral deviation was allowed. The organised-line task succeeded on 75% of trials (18/24); the 6 failures all stemmed from one of two participants (P3, P2) overtaking the target after drifting sideways and re-locking onto the wrong person — an early signal of the target-tracking fragility that the later LineChaser paper addresses with colour-histogram re-identification. Mean reaction time between an extra moving and the blind follower moving was 3.55 s (SD 2.66), significantly slower than the 1.23 s reaction of sighted extras (Welch t-test, p<0.001). SUS mean was 78.3 (grade B+), with five of six participants rating the system 77.5–97.5 and one participant (P2) scoring it 37.5 (F) because holding the heavy phone in front of her was physically uncomfortable. All Likert-scale confidence and comfort measures rose between the pre- and post-study questionnaires. Four participants raised a social-friction concern that would also carry through to LineChaser: it felt awkward to visibly point a smartphone at the person in front.
Relevance
For practitioners, the value of this short paper is as a minimal-viable-system design study: it shows that a blind user can be helped to maintain queue position using nothing more than an off-the-shelf smartphone's depth camera and a three-pattern vibration vocabulary — no building instrumentation, no learned user-specific model, no network. That is a strong baseline to compare richer systems against, and the 94.5% acceptable-position rate is high enough that a simpler product could plausibly ship. The study also identifies three design debts that the fuller LineChaser paper then pays down: (1) target miss-tracking when the user shifts sideways (resolved via colour-histogram re-ID), (2) no support for finding the line's end in the first place (resolved via AR-marker + ARKit localisation on a prepared floor map), and (3) no directional feedback back to the target once lost (partially resolved via audio clock-position). As an artefact in the literature, this paper is most useful when read alongside the CHI '21 LineChaser follow-up; on its own it demonstrates that line-standing assistance is tractable but far from complete, and that the social-acceptance cost of visibly pointing a phone camera is real.
Tags: line standing · pedestrian detection · vibration feedback · visual impairment · orientation and mobility · smartphone assistive technology · blind navigation · late-breaking work