AudioGraf: A Diagram-Reader for the Blind
Andrea R. Kennel · 1996 · Proceedings of the Second Annual ACM Conference on Assistive Technologies (Assets '96) · doi:10.1145/228347.228357
Summary
This paper from ETH Zürich presents AudioGraf, a system that enables blind and visually impaired users to read diagrams through a combination of touch panel input and auditory display. The core problem addressed is that diagrams — widely used in technical reports and papers to represent interrelationships — are inherently visual and inaccessible to blind users. Existing solutions at the time included full-screen tactile displays (expensive), tactile maps printed on special printers (requiring sighted preparation and costly hardware), and audio-only approaches (limited by the non-directional nature of hearing). AudioGraf takes a different approach: the diagram is displayed on a touch-sensitive glass panel, and when the user touches a location, the system auditorily reports what graphic elements are at that position. The system models diagrams as three types of graphic elements: frames (rectangles, ovals with attributes like position, size, color, and shade), text elements (strings with font and position), and connections (lines with endpoints, nodes, arrows, and line attributes). The paper presents a formal model of audio-tactile exploration based on McAdams's model of auditory perception, describing how users cycle through identifying graphic elements, interpreting them against personal knowledge, building a mental map, and searching for gaps or inconsistencies until the diagram is fully understood.
Key findings
Usability tests demonstrated that blind users could successfully read and identify graphic elements in simple diagrams. AudioGraf provides two display levels: a counter view (showing the number and types of elements at the touch point) and an element view (showing detailed attributes of a specific element reached by pressing harder). Users found the small (4mm) and medium (16mm) focus sizes most helpful, while zero focus was too precise and large focus was too imprecise — users who tried large focus instead requested a quantitative description listing all frames and connections as spoken words. The system used both spoken words and auditory symbols (non-speech sounds) to convey information, with connections represented as plucked string sounds and text elements spoken aloud. A key observation was that users who became blind later in life had an easier time interpreting diagrams than congenitally blind users, because they could draw on visual recognition experience. The pressure-based switching between counter and element views was found to be intuitive and easy, though the pressure threshold needed to be set relatively high to prevent accidental view changes.
Relevance
AudioGraf represents an important early approach to a problem that remains largely unsolved: making diagrams and other non-textual graphical content accessible to blind users. While screen readers can convey text, they cannot inherently represent spatial relationships, connections, and visual structures that diagrams encode. The audio-tactile exploration model presented in this paper — where the user actively explores spatial content by touch while receiving auditory feedback — anticipates modern approaches to accessible data visualization and tactile graphics. The distinction between counter view (overview) and element view (detail) parallels the overview-then-detail pattern recommended in current accessibility guidance for complex graphics. For practitioners, the paper underscores that diagram accessibility requires more than alternative text descriptions; spatial and relational information needs interactive exploration methods. The finding that adventitiously blind users performed better than congenitally blind users highlights the importance of considering users' prior visual experience when designing non-visual interfaces.
Tags: blindness and low vision · auditory display · tactile graphics · diagram accessibility · non-visual access · touch panel · sonification · audio-tactile