Automated Sign Language
report
This report provides an overview of a Proof-of-Concept for a signing avatar service. The Proof of Concept was created by TNO in an assignment from the Research and Development department of the Directorate Distribution and Broadcasting of the Netherlands Public Broadcasting Organisation (Directie Distributie en Uitzending, Nederlandse Publieke Omroep), the NPO. In 2012, the NPO asked TNO to investigate the feasibility of a signing service for children’s TV-programs based on a virtual character, a so-called avatar. The NPO believes that such technology could significantly increase the number of programs that can be made more accessible to deaf and hard-of-hearing children.
Based on the current state of technology we concluded that it is not feasible to allow an avatar to sign according to Nederlandse Gebarentaal (NGT, Dutch Sign Language), but that a solution based on Nederlands met Gebaren (NmG, Dutch with Signs) probably is feasible.
Based on the findings we estimate that a fully automated workflow, suitable for integration in a broadcaster workflow, will not be possible within 5 years and probably even longer. On the other hand, the user study shows the need and added value of such a service at the present time. It therefore will be beneficial to further explore semi-automated solutions.
Our main recommendation to the NPO is to continue this research in a European collaborative setting, as many of the remaining key challenges are independent of the respective sign language. Additionally, economies of scale will make avatarbased signing services in the short term much more viable if the underlying technologies can be used across Europe – or even beyond.
Based on the current state of technology we concluded that it is not feasible to allow an avatar to sign according to Nederlandse Gebarentaal (NGT, Dutch Sign Language), but that a solution based on Nederlands met Gebaren (NmG, Dutch with Signs) probably is feasible.
Based on the findings we estimate that a fully automated workflow, suitable for integration in a broadcaster workflow, will not be possible within 5 years and probably even longer. On the other hand, the user study shows the need and added value of such a service at the present time. It therefore will be beneficial to further explore semi-automated solutions.
Our main recommendation to the NPO is to continue this research in a European collaborative setting, as many of the remaining key challenges are independent of the respective sign language. Additionally, economies of scale will make avatarbased signing services in the short term much more viable if the underlying technologies can be used across Europe – or even beyond.
TNO Identifier
505860
Publisher
TNO
Collation
41 p.
Place of publication
Delft