Since one of the goals was to design an extensible system that was easy-to-use, this almost precludes the possibility of manually-programmed analysers. This implies that computer-based learning approaches to solving this problem might be viable. To do so, there are a number of learning methodologies that can be used to build a system capable of recognising gestures.
In general, the task of these learning algorithms is one of classification -- given a set of inputs, commonly known as attributes these systems must decide the type or class of the object based on those attributes. The attributes need not in general be directly physical, and frequently some form of pre-processing is required to provide the attributes actually used by the learning algorithms.