Abstract
Thanks to the evolution of computer science in the 2000s, electronic corpus are developed,allowing researchers to have better bases for sign language research. However, the videos
composing these corpus must be annotated by hand. Thenceforward, it is interesting to
have a solution to accelerate this annotation process by automation. In this memoir, we
will evaluate the feasibility of such a solution.
We first worked for four months at the Laboratory of French Belgian Sign Language
(LSFB-lab) of the University of Namur, where we developed a system to automate annotation
of the videos that compose its corpus. Then we tested our solution on a sample of videos
before comparing our results with the manual annotations.
The results obtained seem to show that a solution which aim to automate the annotation
process of the videos composing the LSFB-lab corpus is not possible under the current
conditions. However, other solutions are possible to accelerate this annotation process : one
based on automation but requiring to record new videos taking into account the limitations
of automation techniques, the other based on collaborative system to keep current videos.
Keywords : sign language - recognition - automation
| Date of Award | 22 Jun 2017 |
|---|---|
| Original language | French |
| Awarding Institution |
|
| Supervisor | Vincent Englebert (President) & Anthony Cleve (Supervisor) |