The Challenge
The VOiCES from a distance challenge 2019 will be focused on benchmarking and further improving state-of-the-art technologies in the area of speaker recognition and automatic speech recognition (ASR) for far-field speech. The challenge will use the VOiCES data for development and evaluation. The challenge will have two tracks for speaker recognition and ASR:
(i) Fixed System - Training data is limited to a subset of VOiCES, a subset of LibriSpeech, SITW, and VoxCeleb
(ii) Open System - Participants can use any external datasets they have access to (private or public)
Participants who complete the challenge (submit their system outputs and system description) will get early access to the VOiCES phase 2 data.
The phase 2 data is an extension of VOiCES, with over 310k audio files recorded in new reverberant environments.
Updates and News
Feb. 27, 2019
- Evaluation data for speaker recognition and automatic speech recognition (ASR) tasks
is now available. Note
system outputs are due by March 4th, 2019 11:59 PM PST.
Jan. 24, 2019
- The official scorer for speaker recognition is now released
- Updated ASR development set is released.
- Updated evaluation plan is released.
Data and Evaluation
Data subsets for challenge tracks are available for download HERE. Make sure to check-out the README.VOICES_2019.txt for more details on what is included in each subset and tips on how to to unzip the data.
The evaluation plan outlines training conditions for fixed and open tracks, details on the development and evaluation data, model performance metrics, and how to submit scores, for both ASR and SID tasks. This document also provides information on how the training and evaluation datasets provided for download are organized, and evaluation rules for the challenge.
Timeline
Jan. 15, 2019 : Release of the evaluation plan and development sets
Feb. 25, 2019 : Evaluation data available
March 4, 2019 : System output submission deadline (11:59 PM PST)
March 11, 2019 : Release of evaluation results
March 15, 2019 : System description submission and release of VOiCES phase II key for the participating teams
March 29, 2019 : Regular paper submission deadline for Interspeech 2019
FAQ’s
You’ve got questions, we’ve got answers.
Additional questions about the challenge? Contact us at voices_poc@sri.com