Modeling and Generating Sign Language as Animated Line Drawings
Frank Godenschweger, Thomas Strothotte · 1998 · Proceedings of the Third International ACM Conference on Assistive Technologies (Assets '98) · doi:10.1145/274497.274513
Summary
This paper from Otto-von-Guericke University of Magdeburg introduces a system for generating sign language as animated line drawings rather than photorealistic 3D renderings. The authors argue that the prevailing approach of creating increasingly realistic virtual humans for sign language synthesis is both computationally expensive and unnecessary — photorealistic quality in virtual signers was not achievable with the technology of the time, and the attempt to approximate it created an uncanny effect. Instead, they propose using line drawings, which have three key advantages: they work with simpler 3D models without losing essential information, the resulting images can be transferred very quickly over networks (important for Internet applications), and they naturally resemble the illustrations used in printed sign language teaching materials. The system decomposes sign language into four components: left hand sign, right hand sign, body movement, and facial expression. Each gesture is composed from these elements and stored in a library with verbal descriptions. Animated signs are built from keyframe gestures with smooth interpolation between them at 15 frames per second. Facial expressions are modelled by mapping cubic B-spline curves onto a fixed 3D head model, representing mouth, eyes, eyebrows, and wrinkles. The authors emphasise that facial expressions are essential to sign language — without them, signs may be misinterpreted or not understood at all.
Key findings
The system's line-drawing approach offered several technical innovations. All processing operated on analytical models (lines and control points) rather than pixel matrices, making computation inexpensive. The line rendering tool allowed significant variation in line quality — width, saturation, sketchiness, and line-end appearance — enabling selective emphasis of different parts of a sign. For example, finger positions could be emphasised with thicker or different-style lines when they were the critical feature of a sign, or facial features could be made more prominent when mouth movements were essential. This ability to direct visual attention through line style variation was considered particularly valuable for teaching applications. The facial expression system used B-spline curves mapped to a head model's surface, with control points in local coordinates, allowing smooth interpolation between different expressions (demonstrated with a 12-frame transition from "surprised" to "smiling"). The system demonstrated German Sign Language signs including "Niemals" (never) and "Baum" (tree). Feedback from sign language teachers, including one who was deaf, was described as very encouraging — the emphasis capability was considered a key element for teaching systems, and the line drawing style was well received as matching printed materials.
Relevance
This paper introduced an important alternative paradigm for sign language synthesis: rather than pursuing ever-more-realistic virtual humans, it used deliberate abstraction to achieve clarity and teachability. The line-drawing approach anticipated a broader principle in accessibility: sometimes less visual detail creates more effective communication. The ability to selectively emphasise parts of a sign through visual style — drawing attention to fingers, face, or body movement as needed — is a pedagogical affordance that photorealistic rendering doesn't naturally provide. For modern sign language technology, the paper's core insights remain relevant: signing avatars used on websites and in apps today still struggle with the balance between realism and clarity, and the principle that abstracted representations can be more effective than realistic ones for learning and comprehension is well supported by subsequent research. The system's suitability for low-bandwidth network transmission also anticipated the need for lightweight sign language content delivery that remains relevant for mobile and developing-world contexts.
Tags: sign language · sign language animation · deafness · computer graphics · non-photorealistic rendering · assistive technology · sign language synthesis