LSF-ANIMAL: A Motion Capture Corpus in French Sign Language Designed for the Animation of Signing Avatars

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 6008–6017 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 6008 LSF-ANIMAL: A Motion Capture Corpus in French Sign Language Designed for the Animation of Signing Avatars Lucie Naert, Caroline Larboulette, Sylvie Gibet IRISA Lab, Université Bretagne Sud Vannes, France {lucie.naert, caroline.larboulette, sylvie.gibet}@univ-ubs.fr lsf.irisa.fr Abstract Signing avatars allow deaf people to access information in their preferred language using an interactive visualization of the sign language spatio-temporal content. However, avatars are often procedurally animated, resulting in robotic and unnatural movements, which are therefore rejected by the community for which they are intended. To overcome this lack of authenticity, solutions in which the avatar is animated from motion capture data are promising. Yet, the initial data set drastically limits the range of signs that the avatar can produce. Therefore, it can be interesting to enrich the initial corpus with new content by editing the captured motions. For this purpose, we collected the LSF-ANIMAL corpus, a French Sign Language (LSF) corpus composed of captured isolated signs and full sentences that can be used both to study LSF features and to generate new signs and utterances. This paper presents the precise definition and content of this corpus, technical considerations relative to the motion capture process (including the marker set definition), the post-processing steps required to obtain data in a standard motion format and the annotation scheme used to label the data. The quality of the corpus with respect to intelligibility, accuracy and realism is perceptually evaluated by 41 participants including native LSF signers. Keywords: Corpus, Motion Capture, French Sign Language, Marker Set, Annotation, Perceptual Evaluation 1. Introduction Virtual characters, or avatars, are a promising solution to convey information in Sign Language (SL). Avatars make it possible to preserve the signer’s anonymity and to gain in interactivity: the speed, appearance and 3D point of view of the signed sequence can easily be changed by the user. In addition, unlike videos, the signed production can poten- tially be edited to create new signed content. However, in order to be accepted by the Deaf, the avatars must be fully animated with natural, realistic and mean- ingful motions. Those motions can be (i) generated from hand-crafted keyframes (Braffort et al., 2007; McDon- ald et al., 2016), (ii) generated procedurally (Kennaway, 2003; Nunnari et al., 2018), or (iii) generated with data- driven techniques through concatenation of pre-captured motions (Gibet et al., 2011) or machine learning pro- cesses (Brock and Nakadai, 2018). In the first case, the quality of the produced animation can be very high but the process of manually creating the keyframes is fastidi- ous and time-consuming. In the second case, the generated motions are often robotic and unrealistic but any sign can be produced as long as it can be described using a SL repre- sentation such as SigML (Kennaway, 2006), or Azee (Fil- hol and Falquet, 2017). In the third case, the avatar’s mo- tions come from human data …

Content truncated to 500 words