International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 12 | Dec 2025
p-ISSN: 2395-0072
www.irjet.net
Bidirectional Sign Language Translator Dr. Poornima B1, Pooja S N2, Poojitha C H3, Prajwala P N4, Preethi S5 1Professor and Head, Information Science and Engineering, Bapuji Institute of Engineering and Technology,
Davangere, affiliated to VTU Belagavi, Karnataka, India.
2Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and Technology,
Karnataka, India
3Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and Technology,
Karnataka, India
4Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and Technology,
Karnataka, India
5Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and Technology,
Karnataka, India ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Sign language is a very important way for deaf
Language (SL) is the primary and most natural form of communication for millions of deaf and hard-of-hearing people worldwide. But there is still a big communication gap between the signing community and the general public (nonsigners), mainly because there aren't many people who are proficient in SL and there aren't many qualified, live human interpreters. For deaf people, this barrier frequently limits their access to education, hinders their ability to advance in their careers, and reduces their social inclusion.
and hard-of-hearing people to talk to each other, but the fact that there aren't many easy-to-use translation systems makes it hard for them to talk to people who don't sign. This project presents a Bidirectional Sign Language Translator system that facilitates effortless two-way communication. In the primary direction (Sign-to-Text/Voice), hand gestures are captured using computer vision techniques (Open CV) and processed via the Media Pipe framework for accurate, realtime landmark detection and extraction. These features are classified by a machine learning model, such as the Random Forest algorithm, to produce readable text and synthesized speech. In the reverse direction (Text/Speech-to-Sign), spoken or written language is processed using Natural Language Processing (NLP) methods and then mapped into structured sign language sequences, which are displayed in a lightweight textual/GIF format for deaf users.
Technological developments, especially in computer vision and machine learning, have made it possible to create automated translation systems as a solution to this pressing issue. In order to enable smooth, two-way communication between sign language and non-sign language users, this project presents a Bidirectional Sign Language Translator system. Sign-to-Text/Voice and Text/Speech-to-Sign are the two separate but integrated modes in which the system functions.
The system emphasizes efficiency, real-time performance, and device adaptability by consciously avoiding reliance on heavy graphical or 3D animation models. This design choice ensures it can be deployed on standard platforms like mobile devices and web applications without significant computational overhead. Preliminary evaluations demonstrate promising results in recognition accuracy and processing speed. By offering a lightweight, hardware-independent solution, the system reduces dependence on human interpreters, promoting inclusivity, accessibility, and independence for the deaf community. Furthermore, its modular architecture makes it scalable and adaptable to different sign languages for broader real-world applications.
Advanced computer vision techniques form the core of the Sign-to-Text/Voice translation module. It uses the Media Pipe framework to extract high-dimensional anatomical landmarks (key points) and reliably and instantly record hand gestures. To precisely map the signs to matching text and synthesized speech, these extracted features are subsequently fed into an effective machine learning classifier, such as the Random Forest algorithm. Importantly, the system is designed for real-time performance and lightweight deployment, purposefully avoiding complicated, resource-intensive graphical or 3D animation models for sign rendering. This design decision guarantees that the solution is workable, scalable, and easily accessible on everyday devices like smartphones and standard web platforms.
Key Words: Sign Language Translation, Media pipe, Random Forest, Gesture Recognition, Bidirectional Translator, Accessibility, Assistive Technology
In contrast, the Text/Speech-to-Sign module converts spoken or written input into structured sign language sequences using Natural Language Processing (NLP) techniques. The deaf user is then shown this output in a straightforward, informative text or GIF format. This
1. INTRODUCTION Effective communication is essential for social integration, human interaction, and obtaining necessary services. Sign
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 481