Focus of Attention Detection and Attention Guidance

An intelligent environment needs to understand at any time what is in the user’s current focus of attention. Some of the main purposes to have this knowledge available include

  • getting a better picture of the context as part of normal user-environment-dialogue
  • being able to proactively make suggestions, give hints or notify of potential problems
  • validating the result of instructions given by the system (anticipation feedback loop).

To detect the focus of attention, all available sources of information should be consulted, in particular sensor sources that reveal physical information about the user, e.g. head orientation or posture. A multi-scale, multi-hypothesis approach is pursued in order to narrow down the focused item when referring to distant objects, applying different sensors depending on their range and accuracy.

In some cases, the system also likes to point the user at specific objects or locations, e.g. as part of a notification or when the user searches for an item. MADMACS investigates different techniques for this attention guidance, which should always follow usability guidelines and respect the user’s privacy. The realization can make use of all output devices available in the cyber-physical environment (CPE).

Together, these features are referred to as attention management. They form another essential part of the dialogue platform used in MADMACS and will be part of a new attention management component. This component works in conjunction with various sensor and output rendering technologies (e.g. the EYEVIUS system) to obtain its knowledge.