WELCOME TO EJ's
HOME PAGE

I am a research fellow at the Centre for Exploration Targeting and an adjunct research fellow at the School of Computer Science & Software Engineering within the University of Western Australia. At the Computer Science school I gained my BSc(Hons) in 1989, MSc in 1992 and PhD in 1997.
Research
My research focus is on achieving automatic two-way translation
between sign and spoken
languages using computer vision
and graphics techniques. The deaf community in Australia uses a
sign language is called Auslan. The translation of English into
Auslan requires robust speech
recognition, facial expression recognition, as well as a
graphical sign
display. The reverse translation that is from Auslan
into English requires the visual recognition of signing gestures.
This research was funded by the Australian Research Council (ARC), and
is led by Professor
Robyn Owens.
The main focus of the current research includes the following:
- Sign Language Visualisation using
the computer graphics technique:
We developed a
sign visualisation system that uses a computer-generated 3D human model
to display Auslan Signs . This visualisation system has a
interactive tutorial interface, namely the Auslan Tuition System that
helps people learn Auslan. The Auslan Tuition System generates
sign motion in real-time and runs on a domestic PC without a specific
harware, and is adaptable to other sign languages. This sytem was
released as a public
domain software downloadable from the Internet since July
2004. The Auslan Tuition
System received a Commendation in the
Innovation Category of the WA Information Technology and Communications
Award 2004.
(Collaborators: Robyn Owens, Jason Wong, Sam Yeates, Nick Lowe, James
Strauss)
- Automated sign recognition
using the computer vision technique:
Our sign recognition system recognises
Auslan phrases by tracking two hands and the face of a signer;
extracting visual features that are invariant to scaling, 2D rotations
and the signing speed; and recognising sign phrases using hidden Markov
models. With a known colloquial grammar, the system achieved over
a 97% recognition rate for sign phrases.
(Collaborators: Robyn Owens, Gareth Lee)
- Automatic lipreading, automatic surveillance and other projects are shown here.
Teaching
Selected Publications
Recent publications can be viewed and downloaded from here.
Useful Links
Home
| Computer Science
| UWA