The ARC/NHMRC funded Thinking Head project has aims that focus around being embodied in the world without having a body. As he likes to say "I'm just a head!" The project aims to produce a portable platform which works with a variety of hardware and has interchangable components joined in a chain from input to output. We propose to make a full Thinking Head chain available freely for research, so that researchers can concentrate on the functionality and modules they are interested in and can further develop.
The fly in this pleasant-seeming ointment is that the Head is handling enormous volumes of video and audio data, that different modules want to do different things with the same data, and that we want to interface with a variety of standard drivers, libraries and applications. For example, speech and video are used for speaker recognition, speech recognition, expression detection/interpretation, and a variety of other higher and lower level processing - e.g. to be able to understand and produces sentences incrementally, or to have a team of experts competing to respond. Vision, may also be used for face tracking, object tracking and gaze tracking, recognizing sources of interruption or potential subjects of conversation.
TMF handles all the technical stuff: concurrent processing, bufffering, archving and synchronization, and allows for central control of media sources, making information available to processes/tasks that need it.
Authors: David M W Powers, Trent Lewis, Martin Luerssen and Richard Leibbrandt
Event: SF08: Embodied Interaction in Mobile, Physical and Virtual Environments Workshop