Real-Time Word Detection in Turkish Sign Language with Deep Learning

Read full text Summarise with VIRAT
Authors: Abdil Karakan, Yüksel Oğuz
Cited by: 0
References: 42

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Communication occurs when people can mutually understand each other. Hearing-impaired people have great difficulties communicating with the people around them. Hearing-impaired individuals can often understand others through lip reading. However, they often have difficulty expressing themselves to others. The use of sign language has not become widespread around the world. Hearing impaired language; Apart from hearing impaired people, the number of people who know is very low. The study aims to detect the 50 most commonly used words of hearing-impaired individuals in hospitals and especially in emergency services, using deep learning. The study is a word-based detection process, not a letter-based one. In the study, a movement was detected, not a single photograph. For the study, a data set was created using videos taken from different angles of 50 words used in hospitals by 100 volunteers. Grayscale conversion, tilt augmentation, blurring, variability enhancement, noise addition, image brightness change, colour vividness change, perspective change, sizing, and position change were added to the photographs that make up the data set. With these additions, the error that may occur due to any distortion that may occur from the camera is minimized. Thus, the accuracy rate in the detection process with images taken from the camera in real-time has been increased. The created data set was run on the YOLOv8 algorithm. The model achieved an average precision of 95.0% and a mean average precision (mAP) of 95.1%. An accuracy rate of 89.40% was achieved in real-world testing.

Fields of Study & Subjects

Hand Gesture Recognition Systems Advanced Neural Network Applications Human Pose and Action Recognition

Publication details

Journal / source
Celal Bayar Üniversitesi Fen Bilimleri Dergisi
Volume
22
Issue
1
Pages
132–141
Publication date
30 March 2026
Language
EN