Ambient Intelligence: Developing the Infrastructure for Building Interfaces

The most common interface to a desktop computer uses the WIMP model. But as computers are becoming more ubiquitous, more tied to a specific device and less bound to the desktop, this model is becoming less suitable. Think of the interface to the newest tv sets that use voice and gesture recognition; we can imagine household devices in the future having similar voice/gesture interfaces.

This topic is concerned with developing the infrastructure for building such interfaces. It will require a language of voice/gesture commands to be defined, with commands that can be interpreted as controls for a range of consumer devices. It will also look at how these commands are recognised using audio and video data; this is particularly difficult as the data is generally not of good quality - for example, the subject will probably not be facing a video camera.