British Sign Language Detection Using Ultra-Wideband Radar Sensing and Residual Neural Network

Saeed, Umer; Shah, Syed Aziz; Ghadi, Yazeed Yasin; Hameed, Hira; Shah, Syed Ikram; Ahmad, Jawad; Abbasi, Qammer H.

Abstract

This study represents a significant advancement in Sign Language Detection (SLD), a crucial tool for enhancing communication and fostering inclusivity among the hearing-impaired community. It innovatively combines radar technology with deep learning techniques to develop a sophisticated, non-invasive SLD system. Traditional SLD methods often rely on cumbersome wearable devices or struggle with environmental inconsistencies. In contrast, this system utilizes the distinctive ability of radar to function effectively across various lighting conditions. The core of this research lies in its application to British Sign Language (BSL) detection, employing advanced neural network architectures for real-time interpretation. A key highlight is the impressive 92% accuracy rate achieved in BSL recognition, utilizing the Residual Neural Network (ResNet) model. This success is attributed to a comprehensive dataset and the strategic adaptation of ResNet for processing radar data. The fusion of radar technology with deep learning in this context not only marks a novel approach in the field but also establishes this research as a foundational contribution to the realm of SLD. Its implications extend beyond technical achievement, offering a more accessible and inclusive communication alternative for the hearing-impaired.

• Journal Article Type: Article
• Acceptance Date: Feb 2, 2024
• Online Publication Date: Feb 15, 2024
• Publication Date: 2024-04
• Deposit Date: Apr 9, 2024
• Journal: IEEE Sensors Journal
• Print ISSN: 1530-437X
• Electronic ISSN: 1558-1748
• Publisher: Institute of Electrical and Electronics Engineers
• Peer Reviewed: Peer Reviewed
• Volume: 24
• Issue: 7
• Pages: 11144-11151
• DOI: https://doi.org/10.1109/jsen.2024.3364389
• Keywords: British sign language, radar sensing, gesture-reading, contactless detection, ResNet
• Public URL: http://researchrepository.napier.ac.uk/Output/3514413