We are pleased to announce the PASCAL ‘CHiME’ Speech Separation and Recognition Challenge.
In 2006 the PASCAL Network funded the first speech separation challenge, addressing the problem of separating and recognising speech artificially mixed with other speech. The best system was able to achieve super-human performance! We are now turning to a more realistic – but more difficult – scenario: recognising speech in the reverberant multisource mixtures that are typical of everyday listening conditions. Specifically, the challenge will employ binaural, distant-microphone recordings made over a period of several weeks in a real family house.
The challenge will be to separate and recognise simple command utterances which have been convolved with a binaural room impulse response and embedded in this continuous background. The challenge is motivated by the demands of real distant-microphone speech recognition applications and has been designed to draw participation from multiple disciplines including signal processing, computational hearing, machine learning and speech recognition. Evaluation will be through speech recognition results but participants will be allowed to submit either separated signals, robust speech features or the outputs of complete recognition systems. We are interested in measuring the performance of both emerging techniques and established approaches.
A full description of the challenge, including details of the source separation and recognition tasks, the noisy speech data sets, and the rules for participation can be found on the Challenge web site.
Results of the Challenge will be presented at a dedicated one-day workshop that will be held as a satellite event of Interspeech 2011 in Florence, Italy. Participants will be invited to submit abstracts or full papers for presentation at this event.
September 2010: Training and development data are available for download
October 2010: Additional tools are available
February 2011: Test data are released
31st March 2011: Submission deadline for the CHiME 2011 workshop.
If you have any questions please do not hesitate to contact the organisers, chime(at)dcs.shef.ac.uk
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)