Automated Translation of Assamese Sign Language through Deep Learning and Computer Vision Techniques
Year: 2025
Type: book-chapter
Abstract: This research introduces a system for recognizing Assamese Sign Language (ASLR) aimed at improving communication for individuals in the Assamese deaf community.The system utilizes advanced computer vision and deep learning methods to accurately identify and convert static hand gestures representing Assamese alphabets into readable text.A custom dataset was created, consisting of images of 10 selected Assamese alphabet signs ("অ", "আ", "ই", "ঈ", "উ", "ও", "এ", "ক", "জ", "ল"), with image capture performed using OpenCV and hand landmarks detected through MediaPipe.A Convolutional Neural Network (CNN) was then trained to classify these gestures, achieving impressive accuracy with near-perfect precision and recall scores.The system seamlessly converts recognized letters into coherent Assamese words, enabling real-time translation.Issues related to the display of non-ASCII characters were addressed by employing Unicode-compatible fonts.The results of the experiments highlight the system's effectiveness in recognizing sign language, offering significant potential for improving accessibility and communication technologies for people with hearing impairments.Future improvements will involve expanding the dataset to include dynamic gestures and complete sentences, as well as implementing the system for real-time use in mobile applications.This work contributes to the ongoing efforts to create an inclusive environment for individuals with hearing impairments in the Assamesespeaking population
Authors: Himangshu Chetia, Bidisha Bhuyan, Madhusmita Bhuyan, Chhaya Prasad, Chandana Dev, Golam Imran Hussain
Institution: Jorhat Institute of Science and Technology
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
SDG: Reduced inequalities
Open Access status: hybrid
OpenAlex ID: W4415763560