Speech & Language
Project: Speaker Diarisation - Who said What, When?
Most existing speech recognition technologies only work with a
single user. If we want to to recognise multiple speakers, such
as in a meeting environment, we must first separate them into
‘who spoke when?’. This task is known as Speaker Diarization.
With this information, we can then recognise what has been
said and attribute it to the correct speaker.
Mark Sinclair
University of
Edinburgh
Acknowledgements
The DEMOfest organising team would like to extend particular thanks
to Susan Craw