In the last few years there has been a lot of interest in applying speech technology in the context of meeting rooms; this seems to have become the next Grand Challenge for speech technology after the Broadcast News domain has been more or less mastered. Meeting rooms provide an interesting range of problems for speech recognition and related technologies. Headset microphones are not convenient and so distant tabletop microphones and microphone arrays are being targetted and room acoustics become a significant problem. The language used in meetings also differs significantly from scripted broadcast news recordings: we see less formal language, interruptions and backchannelling from meeting participants.

In this talk I'll try to give a state-of-the-art overview of various Meeting Room projects around the world and relate my own experiences taking part in the speaker tracking part of the current NIST Meeting Room evaluation.