The 4th CHiME Speech Separation and Recognition Challenge

The 4th CHiME challenge is now complete. There were entries from 15 different teams who presented their systems at the 4th International Workshop on Speech Processing in Everyday Environments (CHiME 2016). Full results and corresponding workshop papers are available on the results page of this site.

Following the success of the 1st, 2nd and 3rd CHiME challenges we are pleased to announce the 4th CHiME Speech Separation and Recognition Challenge (CHiME-4). The new challenge revisits the datasets originally recorded for CHiME-3, i.e., Wall Street Journal corpus sentences spoken by talkers situated in challenging noisy environments recorded using a 6-channel tablet based microphone array. CHiME-4 increases the level of difficulty by constraining the number of microphones available for testing.

All approaches are welcome, whether they are emerging or established, and whether they rely on speech processing, signal separation or machine learning.


The results of CHiME-4 will be presented at a dedicated workshop that will take place on September 13th at Google in San Francisco (345 Spear St) following Interspeech 2016.

The workshop venue is conveniently located about half a mile from the Interspeech conference centre.


CHiME-4 features three tracks depending on the number of microphones available for testing:

Participants are encouraged to submit results for all tracks. Submitting results for one or two tracks is also allowed.

On this web site you will find everything you need to get started, including,

The baselines have been further improved and now score among the best techniques in CHiME-3.





CHiME is supported by