Wexelblat developed a generalised gesture recognition system as part of the work done by AHIG (Advanced Human Interface Group). In his Master's thesis [Wex93], he suggests a hierarchical approach to gesture recognition. In it, he used three Ascension Flock-of-Bird position trackers (one on each hand, and one on the head), together with a CyberGlove on each hand.
Figure 2.14: Wexelblat's system for use within AHIG.
The outline of his system is shown in figure 2.14. We see that the very low level information feeds into the segmenters (which operate separately for each input, and inform the upper level of the system when there is a change in the behaviour of a particular measurement). These are then captured by proto-feature recognisers, and then feature recognisers, which when a particular set of ``needs'' are satisfied for the feature to be recognised, informs the path module, which looks at all the features. It then creates a frame. This output is then integrated temporally (seeing which things happened at the same time, and if so, encapsulate them). This output is then passed to a gestural interpreter.
Although it has not yet been built, it is conceivable that a gestural interpreter which would be able to understand sign language could be designed. The problem, however, would be to design it to be as extensible as possible. It might be necessary to also add proto-features and features which are customised for use in sign language recognition.