Table 1 Summary of related studies.
Reference | Problem | Target class (Input Category) | Algorithm(s) | Limitations | Innovation |
|---|---|---|---|---|---|
Ethiopian Sign Language Recognition Model Using Deep Learning | Fingerspelling & Dynamic Isolated Signs | CNN | Dataset includes few signers, risking bias; segmentation needs a restricted background | Introduced a deep learning- based ESL recognition model handling both static and dynamic signs | |
Real-Time American Sign Language Recognition | Finger Spelling / Character Level | CNN | Kinect requires extra expense to purchase for scaled-up real-world utilization | Introduced a real-time ASL recognition model integrating CNN with depth sensors for improved accuracy | |
Hand Gesture Recognition for Sign Language Using 3DCNN | Dynamic Isolated Sign | 3DCNN | 3DCNN modeling is not robust enough to capture the long-term temporal dependence of the hand gesture signal | Proposed the use of 3DCNN to learn spatiotemporal features of dynamic gestures for sign language recognition | |
A Real-Time Ethiopian Sign Language to Audio Converter | Fingerspelling & Static Signed Words | CNN | Only limited to static signs; non-manual features are not involved | Developed a real-time Ethiopian sign-to-speech converter to aid communcatiion between deaf and non-deaf individuals | |
Amharic Phrase Level Sign Language Recognition Using Deep Learning | Dynamic Continuous Signs | CNN-LSTM | The model is likely to be signer- dependent as it captures signer- specific features | Applied a CNN-LSTM hybrid model to recognize continuous Amharic sign language phrases, capturing spatial and temporal features |
