Call for papers

The REVERB workshop will host the REVERB challenge which addresses the problem of enhancement and recognition of reverberant speech. The REVERB workshop aims at bringing together researchers from a broad range of disciplines to discuss novel and established approaches to handle reverberant speech based on the results of the REVERB challenge. The REVERB challenge participants will present their approaches and results during the REVERB workshop, and we will draw conclusions about the insights gained by the challenge.


To focus the discussions during the workshop, we will only accept papers related to the REVERB challenge. The REVERB challenge participants are invited to submit papers, describing their approaches for reverberant speech enhancement and/or recognition, and evaluated with the REVERB challenge tasks that are briefly described below. To allow authors to fully describe their approaches and provide extensive experimental results, the paper length will be flexible between 4 to 8 pages (including references).

REVERB Challenge Overview

Recently, substantial progress has been made in the field of reverberant speech signal processing, including both single- and multi-channel de-reverberation techniques, and automatic speech recognition (ASR) techniques robust to reverberation. To evaluate state-of-the-art algorithms and draw new insights regarding potential future research directions, we are organizing the REVERB (REverberant Voice Enhancement and Recognition Benchmark) challenge that will provide an opportunity to the researchers in the field to carry out a comprehensive evaluation of their methods based on a common database and on common evaluation metrics. The REVERB challenge is based on the following tasks.
  • Task 1: Enhancement of reverberant speech with single-/multi-channel de-reverberation techniques.
  • Task 2: Robust recognition of reverberant speech.
The data used for the challenge consist of real and simulated 1-, 2-, and 8-channel recordings in reverberant meeting rooms based on the Wall Street Journal Corpus. The data is common to the aforementioned 2 tasks.

Please visit the REVERB challenge webpages for more details.

created by Nuvio | Webdesign