ASLDetect: Arabic sign language detection using ResNet and U-Net like component DOI Creative Commons
Naif Alasmari, Sultan Asiri

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Май 23, 2025

Sign languages are essential for communication among over 430 million deaf and hard-of-hearing individuals worldwide. However, recognizing Arabic Language (ArSL) in real-world settings remains challenging due to issues like background noise, lighting variations, hand occlusions. These limitations hinder the effectiveness of existing systems applications such as assistive technologies education. To tackle these challenges, we propose ASLDetect, a new model ArSL recognition that leverages ResNet feature extraction U-Net-based architecture accurate gesture segmentation. Our method includes preprocessing steps resizing images 64 × pixels, normalization, selective augmentation improve robustness diverse environments. We evaluated ASLDetect on two datasets: ArASL2018, which features plain backgrounds, ArASL2021, more complex On achieved an accuracy 99.35%, surpassing ResNet34 (99.08%), T-SignSys (97.92%), UrSL-CNN (0.98%). For applied transfer learning from our ArASL2018-trained model, significantly improving performance reaching 86.84% accuracy-outperforming (82.5%), (58.98%), (0.49%). results highlight ASLDetect's accuracy, robustness, adaptability.

Язык: Английский

ASLDetect: Arabic sign language detection using ResNet and U-Net like component DOI Creative Commons
Naif Alasmari, Sultan Asiri

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Май 23, 2025

Sign languages are essential for communication among over 430 million deaf and hard-of-hearing individuals worldwide. However, recognizing Arabic Language (ArSL) in real-world settings remains challenging due to issues like background noise, lighting variations, hand occlusions. These limitations hinder the effectiveness of existing systems applications such as assistive technologies education. To tackle these challenges, we propose ASLDetect, a new model ArSL recognition that leverages ResNet feature extraction U-Net-based architecture accurate gesture segmentation. Our method includes preprocessing steps resizing images 64 × pixels, normalization, selective augmentation improve robustness diverse environments. We evaluated ASLDetect on two datasets: ArASL2018, which features plain backgrounds, ArASL2021, more complex On achieved an accuracy 99.35%, surpassing ResNet34 (99.08%), T-SignSys (97.92%), UrSL-CNN (0.98%). For applied transfer learning from our ArASL2018-trained model, significantly improving performance reaching 86.84% accuracy-outperforming (82.5%), (58.98%), (0.49%). results highlight ASLDetect's accuracy, robustness, adaptability.

Язык: Английский

Процитировано

0