A deep learning-based method combines manual and non-manual features for sign language recognition
Haifa Harrouch, Lamia Trabelsi, Maher Jebali, Olfa Gammoudi · 2025
Read A deep learning-based method combines manual and non-manual features for sign language recognition by Haifa Harrouch, Lamia Trabelsi, Maher Jebali, Olfa Gammoudi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-12827390.txt.
Description
Haifa Harrouch and co-authors present a 2025 system that reads sign language from plain video. Most sign data sets hold only color video, with no depth data. Signs also use face and head cues, not just hands, so a good reader must track all of them at once.
The team first estimates the 3D positions of hand joints from the flat video. Those estimates join other streams, such as image traits of the hands and the pose of the head, inside one learning model. On two public sign sets, the system scored above ninety and eighty-eight percent correct. Adding the 3D hand stream gave a clear gain over prior methods.
Catalog record
| CREATOR | Haifa Harrouch, Lamia Trabelsi, Maher Jebali, Olfa Gammoudi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 48 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |