This is perhaps the application that would be easiest to build from
this thesis. A user would put on an instrumented glove and make a sign
that they wanted to know the meaning of. The software would then find
the closest set of matching signs. The user could then examine the
meaning of the word, in English, or perhaps it could be explained in
terms of Auslan itself
. There are
currently a variety of methods for indexing signs, but they are
usually not as intuitive as making the sign itself.
This would be viable, since it uses discrete signs, and a high error
rate is not a large problem, since several candidate signs can be
returned. Furthermore, the algorithms discussed here can both be
expanded to return multiple values.
The only problem with this is that it requires that GRASP be able to adapt to new signers -- ie be able to successfully recognise signs by people other than those on whom it was trained.