Publication Type : Conference Paper
Publisher : Springer Nature Singapore
Source : Lecture Notes in Networks and Systems
Url : https://doi.org/10.1007/978-981-96-5726-1_1
Campus : Coimbatore
School : School of Artificial Intelligence - Coimbatore
Year : 2025
Abstract : This work aims to improve social inclusion and communication for the country’s 300,000+ deaf population through the development of recognition technology for Nepali Sign Language (NSL). Social connection has been limited by NSL’s lack of technological focus in comparison to larger sign languages. To address this problem, we developed a high-accuracy NSL gesture detection system based on the YOLOv8 architecture. On the NSL23 dataset, which comprises 1205 motions in 630 videos, its accuracy was 98.6%. The dataset was then accordingly labeled and preprocessed including augmentation to introduce more variation. Due to the capability of our approach being in real-time, it greatly improves accessibility for NSL users in practicing the actual use of the assisting equipment. This work presents a scalable solution for recognizing NSL, which could serve as the starting point of further research into developing sign language technology, especially for less-resourced languages.
Cite this Research Publication : M. Vignesh, K. Jaidev, T. Taruneshwaran, P. Seetharaman, V. Sowmya, Deep Learning Approach for Nepal Sign Detection, Lecture Notes in Networks and Systems, Springer Nature Singapore, 2025, https://doi.org/10.1007/978-981-96-5726-1_1
