It gives us great pleasure to announce the 6th CHiME Speech Separation and Recognition Challenge (CHiME-6).
The new challenge will revisit the CHiME-5 24-microphone dinner party conversational speech recognition scenario by: - providing an accurate array synchronization script, - introducing a new multichannel diarization track in the line of the DIHARD-II challenge - offering upgraded, state-of-the-art diarization, enhancement, and recognition baselines.
Participants will need to acquire a licence for the CHiME-5 dataset. Licence applications are typically processed within 48 hours. Apply now at: https://licensing.sheffield.ac.uk/i/data/chime5.html Non-commercial license requests are processed within 2-3 days once they have been approved by an authorized representative of your institution. Commercial licenses are also available for companies and are required for challenge participation. Note that licenses acquired before Dec 31, 2018, are no longer valid.
IMPORTANT DATES
Dec, 2019 ? Release of data and baseline systems 10th April, 2020 ? System submission 4th May, 2020 ? CHiME-6 Workshop (satellite of ICASSP 2020) and release of results
ORGANISERS
Shinji Watanabe, Johns Hopkins University, USA Michael Mandel, CUNY, USA Jon Barker, University of Sheffield, UK Emmanuel Vincent, Inria, France
