A UT Arlington computer science
engineering researcher has won two new National Science Foundation grants to
further his work in developing a computer recognition system that
eventually will yield a visual, online American Sign Language dictionary.
Vassilis Athitsos, an assistant professor in the College of Engineering, has now secured more than $1.15 million for research on gesture recognition since joining The University of Texas at Arlington in 2007.
One of the new grants will fund collaborative work among faculty from Boston, Rutgers and Gallaudet universities to organize large collections of sign language video and make them easy to search and download from the web. Gallaudet University in Washington, D.C. is the world's only university in which all programs and services are specifically designed to accommodate deaf and hard of hearing students.
“UT Arlington’s specific part in this project is coming up with a system that allows people to submit videos of signs online,” Athitsos said. “We then identify if and where those signs appear in our videos.”
Athitsos said the goal is to give researchers and the general public better access to these collections of videos.
Athitsos initially became interested in decoding American Sign Language while taking a college course in the subject.
“I was a horrible student; I had to page through a book and look at the signs until I recognized something,” he said. “That is how I thought of building a computer system to help looking up the meaning of signs. My professor at the time told me no one had tried to make a system like that.”
The second grant is a cooperative agreement with ClopiNet, a California consulting company that specializes in pattern recognition, machine learning, statistical data analysis and data mining.
Athitsos said this collaborative project with Isabelle Guyon, the ClopiNet owner, would organize a gesture recognition challenge.
“We will invite participants to produce gesture recognition systems, present them at a conference and have them evaluated live on challenging data sets,” Athitsos said. “It will give all participants a chance to learn from each other as systems are presented. Once the challenge takes place, then we’ll have an expert panel evaluate those systems.”
He said the project’s eventual goal would be to produce better gesture recognition systems, maybe by combining approaches used in the systems presented.
Athitsos’ work is representative of the innovative research under way at The University of Texas at Arlington, a comprehensive research institution of nearly 34,000 students in the heart of North Texas. Visit www.uta.edu to learn more.
The University of Texas at Arlington is an Equal Opportunity and Affirmative Action employer.