...difficult

In many ways, dynamic Bayes nets suffer the same problems as HMMs. In fact, it can be proved that dynamic Bayes nets and hidden Markov models are isomorphic. Dynamic Bayes nets do allow for far more complex state models and provide a much more intuitive mechanism than HMMs for inclusion of background knowledge.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...possible

It is possible to imagine a more complex learning situation, which we might term strong temporal classification, where instead of mapping to 39#39, 35#35 maps to 43#43. In other words, each stream does not map to a single class, but to a sequence of classes. Of course, this makes the learning task much more difficult. It would add the task of segmenting the stream, i.e. deciding which part of the stream is associated with each class label. This is one of the avenues of future research.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...language

Note that recognising sign language as a whole is extremely difficult - to mention three serious difficulties: the use of spatial pronouns (a certain area of space is set to represent an entity); the use of classifiers - signs that are physically descriptive but are not rigidly defined; and improvisation, the creation of new signs based on other signs.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...algorithm

In fact, although not discussed in the model so far, due, perhaps to its impracticality, it is possible to run multiple clustering algorithms as well and use the synthetic events generated by each.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...finger

The little finger is actually important in many sign languages. For example, in Auslan and British Sign Language, a fist with the little finger extended indicates bad. It is also used as a modifier for other signs that indicate badness - e.g. sick is a bad handshape against the body and swear is a bad handshape starting at the mouth and moving away.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...representative

This, of course, makes recognition more difficult. For example, approximately 10 per cent of the signs were two-handed, even though we only had one glove. Several used little finger information.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...HREF="node68.html#figclusterestimateeps">6.4

Note that while the boundary between the shaded and the white area is drawn as an ellipse, this is not the correct shape. The shape (for which a closed form is extremely difficult to compute) is the locus of all values of 87#87 such that 88#88, where 89#89.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...possible

The software is being written in Java to simplify expandability and portability, though it does suffer a performance penalty by doing so. Once completed, the source will be released publicly with the explicit goal of allowing and facilitating others adding their own PEPs, clusterers, feature selectors and learners.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.