Table 4 Results of reviewed works for static image approaches.
Year | Features | Database | Accuracy in (%) |
|---|---|---|---|
201129 | American sign language with Kinect | American sign language | 97 |
20147 | SURF and SIFT | 82.8 | |
20166 | CNN | American sign languages | 80.34 |
201814 | Modified inception model | American sign languages | Average validation:90; Greatest:98 |
201824 | Fusion between RGB and depth image (RBM) | Massey, Fingerspelling A, NYU, ASL fingerspelling of the surrey university | ASL finger spelling A – 98.13 |
20182 | IMU-based glove | Inertial Measurement Units (IMUs), French Sign Language (LSF) | 92.95 |
201931 | YCbCr + SkinMask fusion | custom—1800 images, 20 gesture | Softmax:96.29; SVM:97.28 |
202022 | Random forest, naïve bayes, svm, logistic regression, knn, mlp | ASL, Kaggle32 | KNN: 95.81; ORB & MLP:96.96 |
Proposed method | Multi-headed CNN | American sign language | 98.98 |
