CHiME-2

News - New track 2 baseline system now available

The ASR component of the best performing Track 2 system is now available for download as a baseline implemented using the Kaldi speech recognition toolkit: WSJ0public/CHiME2012-WSJ0-Kaldi_0.03.tar.gz [102 MB] (Note, this is a baseline ASR system and does not include a speech enhancement module.)

The challenge is now officially closed and the results are available here (track 1) and here (track 2). The instructions and data can be found on this site for the benefit of groups wishing to compare their algoriths with those that have been submitted. If citing the challenge please use the following reference,

  • Vincent, E., Barker, J., Watanabe, S., Le Roux, J., Nesta, F. and Matassoni, M. (2013) The second CHiME Speech Separation and Recognition Challenge: Datasets, tasks and baselines In: Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing, Vancouver

    • Organisers

    • Emmanuel Vincent
      INRIA, France
    • Jon Barker
      Univ. of Sheffield, UK
    • Shinji Watanabe &
      Jonathan Le Roux
      MERL, USA
    • Francesco Nesta &
      Marco Matassoni
      FBK-IRST, Italy

    Industrial Board

    • Masami Akamine
      Toshiba
    • Carlos Avendano
      Audience
    • Li Deng
      Microsoft
    • Erik McDermott
      Google
    • Gautham Mysore
      Adobe
    • Atsushi Nakamura
      NTT
    • Peder A. Olsen
      IBM
    • Trausti Thormundsson
      Conexant
    • Daniel Willett
      Nuance
    Supported by the AASP, MLSP and SLTC IEEE Technical Committees

    The 2nd `CHiME' Speech Separation and Recognition Challenge

    One year ago the 2011 PASCAL CHiME Speech Separation and Recognition Challenge considered the problem of recognising speech mixed in two-channel nonstationary noise typical of everyday listening conditions. Following the success of this challenge we are now organising a new challenge that, while keeping the same setting, extends the difficulty along two independent tracks: a larger vocabulary size and a more realistic mixing process that accounts for small head movements made while speaking.

    All approaches are welcome, whether they are emerging or established, and whether they rely on speech processing, signal separation or machine learning. On this web site you will find everything you need to get started, including the general motivation and the details of the small vocabulary and medium vocabulary tracks and, for each track, a number of training and development datasets and baseline recognition and evaluation tools. Some important dates are listed in the schedule.

    The results of the challenge will be unveiled at a dedicated workshop that will be held in conjunction with ICASSP 2013.

    If you have any question or if you are interested in participating, please contact us to receive further updates.

    Acknowledgements This challenge is part of the AASP challenge series. We thank members of the AASP challenge subcommittee and IEEE TC committee chairs for their support and encouragement. We are proud to acknowledge workshop sponsorship from,

    Table of contents