KuSL2023: A standard for Kurdish sign language detection and classification using hand tracking and machine learning
Sign Language Recognition (SLR) plays a vital role in enhancing communication for the deaf and hearing-impaired communities, yet there has been a lack of resources for Kurdish Sign Language (KuSL). To address this, a comprehensive standard for KuSL detection and classification has been introduced. T...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-06-01
|
| Series: | MethodsX |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2215016125002201 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Sign Language Recognition (SLR) plays a vital role in enhancing communication for the deaf and hearing-impaired communities, yet there has been a lack of resources for Kurdish Sign Language (KuSL). To address this, a comprehensive standard for KuSL detection and classification has been introduced. This standard includes the creation of a real-time KuSL recognition dataset, focusing on hand shape classification, composed of 71,400 images derived from merging and refining two key datasets: ASL and ArSL2018. The ArSL2018 dataset, aligned with the Kurdish script, contributed 54,049 images, while the ASL dataset added 78,000 RGB images, representing 34 Kurdish sign categories and capturing a variety of lighting conditions, angles, and backgrounds. Various machine learning models were employed to evaluate system performance. The CNN model achieved an accuracy of 98.22 %, while traditional classifiers such as KNN and LightGBM reached 95.98 % and 96.94 %, respectively, with considerably faster training times. These findings underscore the robustness of the KuSL dataset, which not only delivers high accuracy and efficiency but also sets a new benchmark for advancing Kurdish Sign Language recognition and broader gesture recognition technology. • Provides the first standardized dataset for Kurdish Sign Language recognition using 71,400 annotated images. • Demonstrates high classification accuracy using CNN (98.22 %) and traditional models like KNN and LightGBM. • Enables real-time hand sign recognition and supports the development of assistive technologies for the deaf community. |
|---|---|
| ISSN: | 2215-0161 |