@InProceedings{TeodoroBernDigi:2017:SkCoSe,
author = "Teodoro, Beatriz Tomazela and Bernardes, Jo{\~a}o and
Digiampietri, Luciano Antonio",
affiliation = "USP and USP and USP",
title = "Skin Color Segmentation and Leveshtein Distance Recognition of BSL
Signs in Video",
booktitle = "Proceedings...",
year = "2017",
editor = "Torchelsen, Rafael Piccin and Nascimento, Erickson Rangel do and
Panozzo, Daniele and Liu, Zicheng and Farias, Myl{\`e}ne and
Viera, Thales and Sacht, Leonardo and Ferreira, Nivan and Comba,
Jo{\~a}o Luiz Dihl and Hirata, Nina and Schiavon Porto, Marcelo
and Vital, Creto and Pagot, Christian Azambuja and Petronetto,
Fabiano and Clua, Esteban and Cardeal, Fl{\'a}vio",
organization = "Conference on Graphics, Patterns and Images, 30. (SIBGRAPI)",
publisher = "IEEE Computer Society",
address = "Los Alamitos",
keywords = "sign language recognition, image processing, human skin
segmentation, Brazilian Sign Language, LIBRAS.",
abstract = "Sign language automatic recognition is an important research area
with open challenges that aims to mitigate the obstacles in the
daily lives of people who are deaf or hard of hearing and increase
their integration in the predominantly hearing society in which we
live. This paper implements, evaluates and discusses strategies
for automatic recognition of Brazilian Sign Language (BSL) signs,
which ultimately aims to simplify the communication between deaf
signing in BSL and listeners who do not know this sign language,
accomplished through the processing of digital videos of people
communicating in BSL without the use of colored gloves or data
gloves and sensors or the requirement of high quality recordings
in laboratories with controlled backgrounds or lighting. An
approach divided in several stages was developed and all stages of
the proposed system can be considered contributions for future
works in sign language recognition or those involving image
processing, human skin segmentation, object tracking etc. For the
skin color based segmentation stage, in particular, several
techniques were implemented and compared and the strategy used for
sign recognition, exploring the Leveshtein distance and a voting
scheme with a binary classifier, is unusual in this area and
showed good results. From the original 600 samples of 30 words,
chosen for frequency of use and superposition of sign elements to
make recognition more complex, the system was able to correctly
segment 422 (70%) signs, for which it reached 100% accuracy in
recognition using our strategy. This sign database with 600
samples in video of the chosen 30 word vocabulary is another of
this works contributions and is available upon request to the
authors.",
conference-location = "Niter{\'o}i, RJ, Brazil",
conference-year = "17-20 Oct. 2017",
doi = "10.1109/SIBGRAPI.2017.19",
url = "http://dx.doi.org/10.1109/SIBGRAPI.2017.19",
language = "en",
ibi = "8JMKD3MGPAW/3PF84FS",
url = "http://urlib.net/ibi/8JMKD3MGPAW/3PF84FS",
targetfile = "Skin_Color_Segmentation_and_Leveshtein_Distance.pdf",
urlaccessdate = "2024, May 02"
}