TY - GEN
T1 - Activity detection in conversational sign language video for mobile telecommunication
AU - Cherniavsky, Neva
AU - Ladner, Richard E.
AU - Riskin, Eve A.
PY - 2008
Y1 - 2008
N2 - The goal of the MobileASL project is to increase accessibility by making the mobile telecommunications network available to the signing Deaf community. Video cell phones enable Deaf users to communicate in their native language, American Sign Language (ASL). However, encoding and transmission of real-time video over cell phones is a powerintensive task that can quickly drain the battery. By recognizing activity in the conversational video, we can drop the frame rate during less important segments without significantly harming intelligibility, thus reducing the computational burden. This recognition must take place from video in real-time on a cell phone processor, on users that wear no special clothing. In this work, we quantify the power savings from dropping the frame rate during less important segments of the conversation. We then describe our technique for recognition, which uses simple features we obtain "for free" from the encoder. We take advantage of the conversational aspect of the video by using features from both sides of the conversation. We show that our technique results in high levels of recognition compared to a baseline method.
AB - The goal of the MobileASL project is to increase accessibility by making the mobile telecommunications network available to the signing Deaf community. Video cell phones enable Deaf users to communicate in their native language, American Sign Language (ASL). However, encoding and transmission of real-time video over cell phones is a powerintensive task that can quickly drain the battery. By recognizing activity in the conversational video, we can drop the frame rate during less important segments without significantly harming intelligibility, thus reducing the computational burden. This recognition must take place from video in real-time on a cell phone processor, on users that wear no special clothing. In this work, we quantify the power savings from dropping the frame rate during less important segments of the conversation. We then describe our technique for recognition, which uses simple features we obtain "for free" from the encoder. We take advantage of the conversational aspect of the video by using features from both sides of the conversation. We show that our technique results in high levels of recognition compared to a baseline method.
UR - http://www.scopus.com/inward/record.url?scp=67650665571&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=67650665571&partnerID=8YFLogxK
U2 - 10.1109/AFGR.2008.4813363
DO - 10.1109/AFGR.2008.4813363
M3 - Conference contribution
AN - SCOPUS:67650665571
SN - 9781424421541
T3 - 2008 8th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2008
BT - 2008 8th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2008
T2 - 2008 8th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2008
Y2 - 17 September 2008 through 19 September 2008
ER -