Manual annotation is an expensive and time consuming task partly due to the high number of linguistic channels that usually compose sign language data. In this paper, we propose to automatize the annotation of sign language motion capture data by processing each channel separately. Motion features (such as distances between joints or facial descriptors) that take advantage of the 3D nature of motion capture data and the specificity of the channel are computed in order to (i) segment and (ii) label the sign language data. Two methods of automatic annotation of French Sign Language utterances using similar processes are developed. The first one describes the automatic annotation of thirty-two hand configurations while the second method describes the annotation of facial expressions using a closed vocabulary of seven expressions. Results for the two methods are then presented and discussed.
Keywords
Computer recognition of sign language and steps towards automatic annotation
@inproceedings{naert:18014:sign-lang:lrec,
author = {Naert, Lucie and Reverdy, Cl{\'e}ment and Larboulette, Caroline and Gibet, Sylvie},
title = {Per Channel Automatic Annotation of Sign Language Motion Capture Data},
pages = {139--146},
editor = {Bono, Mayumi and Efthimiou, Eleni and Fotinea, Stavroula-Evita and Hanke, Thomas and Hochgesang, Julie A. and Kristoffersen, Jette and Mesch, Johanna and Osugi, Yutaka},
booktitle = {Proceedings of the {LREC2018} 8th Workshop on the Representation and Processing of Sign Languages: Involving the Language Community},
maintitle = {11th International Conference on Language Resources and Evaluation ({LREC} 2018)},
publisher = {{European Language Resources Association (ELRA)}},
address = {Miyazaki, Japan},
day = {12},
month = may,
year = {2018},
isbn = {979-10-95546-01-6},
language = {english},
url = {https://www.sign-lang.uni-hamburg.de/lrec/pub/18014.pdf}
}