MobileASL

1
MobileASL Dane Barney Anna Cavender Neva
Cherniavsky Frank Ciaramello Jessica
DeWitt Loren Merritt Rahul VanamEve
Riskin Sheila Hemami Richard Ladner
University of Washington Cornell
University http//www.cs.washington.edu/research/
MobileASL
User Study
Objective
Deaf people can use mobile videos phones to
communicate in their own language American Sign
Language (ASL).

• What are the effects of encoding face regions
  with higher rates?
• Previous work Receivers of sign look at the face
  of the signer 95 of the time.

Our Goal Create video compression techniques
for ASL.
No ROI
Some ROI
Even more ROI
Skin Detection
Fast Encoding

• Skin detection algorithms can determine which
  macro-blocks contains skin pixels. We could then
  encode those regions at higher rates.

We are choosing encoding parameters that result
in minimum encoding time and distortion.