Abstract
Development of automatic recognition systems for American Sign Language (ASL) needs a comprehensive database that provides a range of signed material under controlled and less-controlled lighting conditions. The database we created contains (a) handshapes in isolation and in single signs, (b) the American fingerspelling alphabet, (c) numbers, (d) movement in single signs, and (e) examples of short discourse narratives for testing sign recognition in connected linguistic contexts.
All of these are produced by 14 fluent Deaf ASL signers under controlled lighting conditions in a professional studio. All except the short narratives are also produced in less than superior lighting conditions.
These data can provide the recognition algorithm developer with the opportunity to move from simple recognition situations in the best of circumstances to more complex recognition situations with challenging lighting situations.
The database was collected with support from the National Science Foundation Linguistics Program under Grant No. 99-05848 and 0414953.
Date of this Version
September 2006