|
| Ref | Year | Technique | Accuracy | Language | Dataset | Dataset size |
|
| [21] | 2023 | CNN | 98.8% top-1 accuracy | Arabic | ArSL2018 | 54,049 images |
| [22] | 2022 | CNN and RNN | Recognition accuracy of 94.46% | Arabic | ArSL2018 | 54,049 images |
| [23] | 2022 | Transfer learning (TL) and RNN | Recognition accuracy of 93.4% | Arabic | New collected data | 2000 videos |
| [24] | 2020 | VGG16 and ResNet152 | 99.6% for ResNet152 and 99.4% for VGG | Arabic | ArSL2018 | 54,049 images |
| [20] | 2020 | CNN, MLP, HMM, RNN, KNN, LDA, SVM, ANN, and more | DL classifiers attained the best performance as compared to other classifiers | Arabic | — | — |
| [25] | 2020 | “Deep learning model” | 90% | Arabic | Raw sign language images that are captured using a camera | 3875 images |
| [26] | 2020 | “RF, SVM, and KNN” | Accuracy value of 83% by using SVM | Arabic | Collected from 10 people who understand sign and ArSL | Dataset of 222 observations |
| [27] | 2020 | “Neutrosophic technique and fuzzy c-means” | 91% | Arabic | Dataset used was provided by “Al-Amal Institute Damietta for deaf students” | 300 images |
| [28] | 2020 | “DeepLabv3C” based on ResNet50 | 89.5% | Arabic | ArSL database | 23 isolated Arabic word signs performed by 3 different users |
| [29] | 2020 | LSTM layer-based NN | 89% for one-handed gestures and 96% for hand gestures | Arabic | Common ArSl dataset | 44 signs were used, 29 of them are one hand gestures, and 15 are two hand gestures |
| [30] | 2020 | Android application “Smart Gloves” | 90% | Arabic | They rely on glove movement gestures | — |
| [31] | 2019 | CNN, KNN, and SVM | 90.02% | Arabic | Collected from Ibn Zohr University | 5839 images |
| [32] | 2017 | “3D CNN,” RNN, KNN, LDA, SVM, ANN, and more | 85% | Arabic | Source not found | 200 videos |
|
