Remind me what I promised to do

Hosting University

Dublin City University

Overview

The amount of time people spend in business meetings has given rise to the by-now well known office joke that calling or attending a meeting constitutes a practical alternative to real work! In fact, meetings are an important aspect of our interaction with others in the workplace and a necessary evil of the knowledge economy. Sometimes, however, we may only have a relatively peripheral interest in a meeting, only being concerned with part of the agenda. If we miss a meeting, typically minutes are available that document the outcomes, although usually this only outlines the decisions taken and doesn't capture the ebb and flow of discussion that led to these decisions. Trying to keep track of when, where and most importantly why decisions were taken can be difficult, making carrying out the associated tasks assigned difficult, particularly if you weren't present at the meeting.

The goal of this project is to instrument a meeting room using a variety of sensors and to use these to record data that is synchronised with the associated meeting documentation (e.g. slides, documents). The sensors we have available are:

* item a stereo camera -- two visible spectrum cameras aligned and pointing at the same scene that can be used to measure depth based on comparing the difference between what each camera ``sees"
* item an infrared camera -- a camera that images emitted heat (e.g. a person's head shows up as a bright pixel patch)
* item an audio sensor network -- multiple microphones positioned in various different locations around the room

Relevance to Host Laboratory

Existing work in the Adaptive Information Cluster has already looked at aligning all these data sources. So for example, based on camera calibration we can image the scene so that at every pixel location we have colour, depth and infrared information. This opens up the possibility of performing all kinds of interesting analysis, such as determining who is writing on a white board and automatically capturing what is written.

Given a meeting record, it then becomes possible to mine this in order to structure it for functionalities such as off-line playback or browsing thereby allowing people to catch up on missed meetings, dipping in and out and only viewing the relevant sections. Indexing can be based on detecting important meeting activities such as the presentations made, questions asked and the discussions that take place. It also opens the door to very interesting higher-level research questions such as determining who the real-decision maker is in an meeting and characterising the other roles assumed by participants. Furthermore, this analysis could also be used to control a virtual camera to focus on who is speaking/presenting for people viewing the meeting remotely, thereby enhancing the experience of remote participation.

In actual fact, equipping a meeting room with sensors in this way and capturing and understanding the activities that transpire is only one point example of a much broader research agenda addressing smart spaces that has applications across many areas such as care for the elderly, safety and security.

Supervisor

Dr. Noel O'Connor     

Students who have worked on this project: