"Scientific understanding and vision-based technological development for continuous sign language recognition and translation"
System Specifications
The specifications of the system are summarized in the next points:
- Multimodal system: Information extracted from both hands, facial expression, head/body position.
- More natural: non-invasive system (no globes and no markers).
- Robust and self-adaptive to the changing ambient conditions: background, light conditions and signer’s clothing.
- Signer-independent.
- Contextual translation: continuous sign language translation within a context, not merely identifying isolated signs; it will take into account where the signs are made within the signing space.
- Multilingual: the main targeted language is the Sign Language of The Netherlands, although the system will be test also with other Sign Languages (in function of the corpora available).
- Modeling of the Spatial Reference Handling: references in signing space occur quite often to refer to previously deposited objects in the virtual signing space.
- Software integration: the different software modules will run under just one application; a graphical user interface (GUI) will control and monitor the translation process.
- It will not be a real time demonstrator: real time factor around 20 for translating NGT.