Computer Generated 3-Dimensional Models of Manual Alphabet Handshapes for the World Wide Web
Sarah Geitz, Timothy Hanson, Stephen Maher · 1996 · Proceedings of the Second Annual ACM Conference on Assistive Technologies (Assets '96) · doi:10.1145/228347.228353
Summary
This paper from Gallaudet University and NASA Goddard Space Flight Center presents a web-based teaching tool for learning American Sign Language (ASL) fingerspelling through interactive 3D computer models. The authors created VRML (Virtual Reality Modeling Language) representations of the 26 ASL manual alphabet handshapes that users could view, rotate, and zoom in any web browser equipped with a VRML viewer. The project addressed a significant barrier in sign language education: the difficulty of learning accurate hand shapes from static 2D images in books or even video, since subtle differences in finger and hand position are critical to correct signing but hard to perceive from a single viewpoint. The 3D models were generated using a CyberGlove — a virtual reality data glove with 18 or 22 resistive strain gauges measuring finger and wrist joint angles — combined with an Ascension Flock of Birds magnetic tracker for wrist position and orientation. Custom software on a Silicon Graphics workstation captured these measurements in real time and produced VRML files. Because the CyberGlove lacked sufficient fidelity for all subtle finger positions, manual tuning of the VRML files was necessary. For letters requiring movement (like J and Z), the authors used VRML 1.0 level-of-detail (LOD) capability with orthographic projection to simulate animation through key positions at different distances.
Key findings
The project successfully produced a complete set of interactive 3D handshape models for all 26 letters of the ASL manual alphabet, accessible via standard web browsers with VRML plugins. The 3D models eliminated the ambiguities inherent in 2D representations by allowing learners to view signs from any angle and distance. The authors provided both high-resolution (400K) and low-resolution (65K) versions to accommodate different connection speeds, with the low-resolution models still conveying hand and finger formations sufficiently. The paper also documents the challenges of the approach: the CyberGlove hardware was insufficient to capture all subtle finger positions accurately, VRML 1.0 did not support true animation (requiring the LOD workaround for dynamic letters), and the system only captured one hand at a time despite ASL requiring two-handed signs for full communication. The authors noted the important observation that visual observation skills for sign language are not well developed in most hearing adults, making close-up 3D inspection of handshapes particularly valuable for learners.
Relevance
This paper represents an early and creative application of web-based 3D technology to sign language education, a domain where the limitations of 2D media create real learning barriers. The core insight — that sign language handshapes are inherently three-dimensional and require 3D representation for effective learning — remains valid today and has been addressed by modern avatar-based sign language tools and motion-capture systems. For accessibility practitioners, the paper illustrates how emerging web technologies can be applied to Deaf education and highlights the ongoing challenge of creating effective digital sign language learning resources. The work also demonstrates an early example of leveraging the web as a platform for distributing accessible educational content to geographically isolated learners, particularly those in rural areas without access to skilled sign language tutors. While the specific technology (VRML) is obsolete, the approach foreshadows current WebXR and 3D web standards being explored for sign language instruction.
Tags: sign language · deaf accessibility · fingerspelling · virtual reality · VRML · 3D modeling · e-learning · American Sign Language
Standards referenced: VRML 1.0