| Since speech is
one of the most intuitive communication methods, it
has been receiving significant commericial and
research attention. The LML is currently
involved with several advanced speech
recognition and synthesis projects. Our
research attempts to
expand the current limits with respect to speaker
independence and noise tolerance.
One focussed project to be able to control
multimedia gear such as CD players via voice
commands. In this case, we are striving
for speaker independent voice recognition. Furthermore,
the system should be resistant to the music
from the speakers.
On the creative side, we are designing a system
for interactive "jam" sessions using MIDI. The
goal is to be able to give voice commands to
the MIDI instruments interactively - setting the
beat; switching instruments; changing the output
of each instrument, etc.
In the virtual human project, we are giving the
virtual receptionist the ability to have
conversations with visitors to the building. The
focus is on speaker independent dialogs where
the visitors do not need to train the system beforehand.
Our goal is to have the virtual human be able
to give information about people and events and
even be entertaining.