Connectionists: Last call: PASCAL CHiME Speech Separation and Recognition Challenge

[Emmanuel Vincent]

       ----------------------------------------------

  PASCAL CHiME Speech Separation and Recognition Challenge

                  Deadline: April 14, 2011
        Workshop: September 1, 2011, Florence, Italy

   http://www.dcs.shef.ac.uk/spandh/chime/challenge.html

       ----------------------------------------------

News:

* Test data: The final test data has now been released and can be 
downloaded from,

http://www.dcs.shef.ac.uk/spandh/chime/PCC/datasets.html

The test set has exactly the same format as the development test set, 
but uses a different selection of 600 Grid utterances mixed with the 
CHiME background at a different set of temporal locations within the 
corpus. The SNRs range from -6 to +9 dB. The utterances are virtually 
positioned at the same spatial location, i.e. directly in front of the 
binaural manikin at a distance of 2 metres, but a different instance of 
the room impulse response has been used.  As before,  the data is 
available at either 16 or 48 kHz, and as either pre-segmented utterances 
or continuous audio. An annotation file has been provided so that 
utterances can be located within the continuous audio. Feel free to use 
the continuous audio and the annotation to determine the acoustic 
context, e.g. for systems that are employing background noise modelling etc.

The existing HTK-based evaluation tools can be applied to the test data 
without modification -- just use the command line argument to point the 
script at the right directory. The tools and documentation are available 
here,

http://www.dcs.shef.ac.uk/spandh/chime/PCC/evaluation.html

* Special Issue:  The Elsevier Journal Computer Speech and Language has 
agreed to publish a Special Issue on the theme of Speech Separation and 
Recognition in Multisource Environments. CHiME challenge participants 
will be invited to submit extended versions of their workshop papers.

* CHiME Workshop: The workshop will take place at the Interspeech venue 
on the day after the main conference has finished, 1st September. The 
workshop will feature a session on the PASCAL CHiME speech separation 
challenge but will also be accepting papers on the general theme of 
source separation and recognition. A call for papers has been issued at
http://www.dcs.shef.ac.uk/spandh/chime/workshop/.

If you require any support with the challenge or have any questions 
please do not hesitate to contact us (chime at dcs.shef.ac.uk). It would 
also be helpful to us at this stage if you could confirm your interest 
in participating.

Best regards,

Jon Barker (University of Sheffield, UK)
Emmanuel Vincent (INRIA Rennes, France)
Ning Ma (University of Sheffield, UK)
Heidi Christensen (University of Sheffield, UK)
Phil Green, (University of Sheffield, UK)