SIGN BRIDGE: Sign Language Translator

DOI

10.2991/978-94-6239-650-0_9How to use a DOI?

Keywords

Mediapipe; ONNXRuntime; iSign CSLRT

Abstract

This work presents SIGNBRIDGE, a browser executed system for real-time recognition of sign gestures using MediaPipe Holistic landmarks combined with a lightweight ONNXRuntime encoder. The model was prepared using the publicly available iSign CSLRT dataset hosted on Hugging Face, containing roughly 7,050 labelled gesture recordings. For the prototype stage, a filtered set of 500 landmark sequences was used to train and validate the encoder. Each sample consists of standardised pose, hand, and facial landmarks arranged into a fixed-length temporal window. The encoder was trained for 50 epochs using a metric-learning objective to pull together embeddings of identical gestures and push apart unrelated ones. After training, gesture-level prototypes were generated by averaging embeddings for each gesture class, enabling prototype-based retrieval during inference. The ONNX-exported encoder processes live browser landmarks and selects the nearest prototype using cosine similarity. To enhance stability, idle/background handling, similarity thresholding, and temporal smoothing through voting buffers were integrated. Tests on the curated dataset confirmed accurate recognition of the trained gesture set, consistent idle suppression, and reduced repeated outputs. The study demonstrates the feasibility of deploying full sign token recognition directly within web browsers, offering a reproducible foundation for future accessible communication systems and gesture-driven interfaces.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Abhijeet Chaurasia
AU  - Aditya Chandgaonkar
AU  - Aayush Mankar
AU  - Ravi Biradar
PY  - 2026
DA  - 2026/04/20
TI  - SIGN BRIDGE: Sign Language Translator
BT  - Proceedings of the Conference on Technologies for Future Cities (CTFC 2025)
PB  - Atlantis Press
SP  - 120
EP  - 129
SN  - 3005-155X
UR  - https://doi.org/10.2991/978-94-6239-650-0_9
DO  - 10.2991/978-94-6239-650-0_9
ID  - Chaurasia2026
ER  -