Assessment of Semantic Taxonomies for Blind Indoor Navigation Based on a Shopping Center Use Case
J. Eduardo Pérez, Myriam Arrue, Masatomo Kobayashi, Hironobu Takagi, Chieko Asakawa · 2017 · Proceedings of the 14th International Web for All Conference (W4A) · doi:10.1145/3058555.3058575
Summary
This paper investigates what environmental information is needed to support effective indoor navigation for visually impaired people, and whether existing geographic data specifications provide adequate semantic vocabularies for this purpose. The researchers surveyed three real-world semantic taxonomy specifications: OpenStreetMap (OSM), the Wayfindr open standard for audio-based wayfinding, and a Japanese government specification from the Ministry of Land, Infrastructure, Transport and Tourism (JMLITT). Each specification categorizes indoor environmental elements — pathways, doorways, elevators, escalators, stairs, public toilets, buildings, and rooms — with varying levels of detail about accessibility-relevant features such as tactile paving availability, handrail locations, braille support, and wheelchair accessibility. The survey revealed substantial overlap across the three specifications despite their different scopes and origins, suggesting emerging consensus on what environmental data matters for accessible navigation. To validate these findings, the researchers created a comprehensive semantic dataset for a large shopping center in Japan comprising 98 stores across 14 floors in three connected buildings. They then used this data to enhance the NavCog smartphone navigation tool, which provides turn-by-turn voice guidance with real-time localization. A navigation experiment with 9 visually impaired participants tested the effectiveness of audio messages enriched with semantic environmental information.
Key findings
Participants gave an overall positive assessment (71.1% positive responses) of the usefulness of the semantic-enriched audio navigation messages. Elevator navigation messages were rated most useful (mean 5.89/7, SD 1.27), while obstacle announcements were rated least useful (mean 4.33/7, SD 1.94). Among white cane users who were totally blind, tactile paving messages were the most valued (mean 6.0). Notably, the one guide dog user suggested turning off obstacle and tactile paving messages entirely, as these were unnecessary and confusing for her navigation style — highlighting that accessible navigation must be customizable to individual needs and mobility aids. Prior smartphone and voice navigation app experience significantly influenced ratings: experienced users rated messages about elevator buttons and navigation approximately 2 points higher than inexperienced users. Obstacle messages proved challenging because obstacles like chairs were difficult to position precisely, leading to localization errors and mixed participant responses. Store name announcements received generally positive ratings (66.7% positive) despite containing only names, suggesting that richer store information could be highly valuable.
Relevance
This research has direct implications for anyone building indoor navigation systems, smart building infrastructure, or location-based accessibility services. The comprehensive taxonomy comparison (Table 1) serves as a practical reference for what environmental data should be collected and tagged to support accessible indoor wayfinding. A key finding for practitioners is that one-size-fits-all navigation does not work — guide dog users, white cane users, people with residual vision, and totally blind individuals all have distinct information needs. Navigation systems must support customizable message types rather than broadcasting all available information. The work also highlights a significant infrastructure challenge: creating and maintaining detailed semantic maps of indoor environments is labor-intensive, and strategies for sourcing this data (from building owners, crowdsourcing, or automated methods) remain an open problem critical to scaling accessible indoor navigation.
Tags: indoor navigation · blindness · wayfinding · location-based services · semantic taxonomy · assistive technology · audio navigation