CHiME

Resources for ASR evaluation in everyday environmets


Learn more

About

The CHiME Challenges

The CHiME challenges are a series of automatic speech recognition evaluations targeting distant multiple microphone speech recognition in everyday listening environments.

There have been four editions of the CHiME challenge to date.

  • CHiME 1 considers binaurally recorded audio in a domestic setting. It employs a small vocabulary ASR task and a static speaker.
  • CHiME 2 revisits CHiME 1 but with a moving speaker and a medium vocabulary ASR task.
  • CHiME 3 considers speech being recorded using a 6 channel tablet device in a number of indoor and outdoor urban environments.
  • CHiME 4 is an extension of CHiME 3 with a 1, 2 and 6 channel setting.

Results of these challenges have been published at a series of CHiME workshop events.

This website is a convenient one-stop location providing information on how to obtain CHiME datasets and baseline system software. You will also find links to the individual CHiME challenge and CHiME workshop websites.


Browse the datasets

CHiME 1

CHiME 1 considers binaurally recorded audio in a domestic setting. It employs a small vocabulary ASR task and a static speaker.

Obtaining the CHiME 1 data
The CHiME 1 dataset is available from the 1st CHiME Challenge website.
Publications using CHiME 1
Outputs of the 1st CHiME challenge were published at the CHiME 2011 International Workshop
If using CHiME 1 please cite the following paper
Barker, J., Vincent, E., Ma, N., Christensen, C. and Green, P., ``The PASCAL CHiME speech separation and recognition challenge,'' Computer Speech and Language, 27(3):621-633, 2013

CHiME 2

CHiME 2 revisits CHiME 1 but with a moving speaker and a medium vocabulary ASR task.

Obtaining the CHiME 2 data
The CHiME 2 dataset is available from the 2nd CHiME Challenge website.
Publications using CHiME 2
Outputs of the 2nd CHiME challenge were published at the CHiME 2013 International Workshop
If using CHiME 2 please cite the following paper
Vincent, E., Barker, J., Watanabe, S., Le Roux, J., Nesta, F. and Matassoni, M., ``The second CHiME Speech Separation and Recognition Challenge: Datasets, tasks and baselines,'' Proc. IEEE ICASSP, Vancouver, 2013

CHiME 3

CHiME 3 considers speech being recorded using a 6 channel tablet device in a number of indoor and outdoor urban environments.

Obtaining the CHiME 3 data
The CHiME 3 dataset is available to WSJ license holders. Details on the 3rd CHiME Challenge website.
Publications using CHiME 3
Outputs of the 3rd CHiME challenge were published at the IEEE ASRU 2015 Workshop
If using CHiME 3 please cite the following paper
Jon Barker, Ricard Marxer, Emmanuel Vincent, and Shinji Watanabe, ``The Third ‘CHIME’ Speech Separation and Recognition Challenge: Analysis and Outcomes'', Computer Speech and Language, to appear

CHiME 4

CHiME 4 is an extension of CHiME 3 with a 1, 2 and 6 channel setting.

Obtaining the CHiME 4 data
The CHiME 4 dataset is available to WSJ license holders. Details on the 4th CHiME Challenge website.
Publications using CHiME 4
Outputs of the 4th CHiME challenge were published at the CHiME 2016 International Workshop
If using the CHiME 4 set up please cite the following two papers
  • Emmanuel Vincent, Shinji Watanabe, Aditya Arie Nugraha, Jon Barker, and Ricard Marxer, ``An analysis of environment, microphone and data simulation mismatches in robust speech recognition'', Computer Speech and Language, to appear.
  • Jon Barker, Ricard Marxer, Emmanuel Vincent, and Shinji Watanabe, ``The Third ‘CHIME’ Speech Separation and Recognition Challenge: Analysis and Outcomes'', Computer Speech and Language, to appear

References

Details of all CHiME challenges and datasets can be found in the following publications

  • Christensen, H., Barker, J., Ma, N. and Green, P., ``The CHiME corpus: a resource and a challenge for Computational Hearing in Multisoure Environments,'' INTERSPEECH'10, Makuhari, Japan, September 2010.
  • Barker, J., Vincent, E., Ma, N., Christensen, C. and Green, P., ``The PASCAL CHiME speech separation and recognition challenge,'' Computer Speech and Language, 27(3):621-633, 2013
  • Vincent, E., Barker, J., Watanabe, S., Le Roux, J., Nesta, F. and Matassoni, M., ``The second CHiME speech separation and recognition challenge: Datasets, tasks and baselines,'' Proc. IEEE ICASSP, Vancouver, 2013
  • Vincent, E., Barker, J., Watanabe, S., Le Roux, J., Nesta, F. and Matassoni, M., ``The second CHiME speech separation and recognition challenge: an overview of challenge systems and outcomes,'' Proc. IEEE ASRU, Olomouc, Czech Republic, 2013
  • Jon Barker, Ricard Marxer, Emmanuel Vincent, and Shinji Watanabe, ``The third CHIME speech separation and recognition challenge: analysis and outcomes'', Computer Speech and Language, to appear
  • Emmanuel Vincent, Shinji Watanabe, Aditya Arie Nugraha, Jon Barker, and Ricard Marxer, ``An analysis of environment, microphone and data simulation mismatches in robust speech recognition'', Computer Speech and Language, to appear.

Download bib

Sign Up

If you would like to receive occasional email updates about future CHiME datasets, challenges or workshops please complete the form below.

Join the CHiME mailing list

(Email addresses will be kept in confidence and not shared with any 3rd parties.)

Or if you have any questions please feel free to contact us by emailing chimechallenge@gmail.com

Sponsors

We would like to thank all the organisations who have given their support to CHiME.

Google logo
ISCA logo
LDC logo


Audience logo
Adobe logo
Conexant logo


EPSRC logo
MERL logo
ROSP sig logo


University of Sheffield logo
Pascal logo
Inria logo