The reverb challenge: A common evaluation framework for dereverberation and recognition of reverberant speech

By Keisuke Kinoshita, Marc Delcroix, Takuya Yoshioka, Tomohiro Nakatani, Armin Sehr, Walter Kellermann, Roland Maas. 2013

431 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Evaluation & Benchmarks, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Evaluation & Benchmarks

In Speech & Audio

In Signal Processing

Citations in the corpus

Light Gated Recurrent Units for Speech Recognition (2018, methodology): “of non-stationary noise and acoustic reverberation [14]–[22]. The development of robust ASR has been recently fostered by the great success of some international challenges such as CHiME [23], REVERB [24] and ASpIRE [25], which were also extremely useful to establish common evaluation frameworks amo”
The DIRHA-ENGLISH corpus and related tasks for distant-speech recognition in domestic environments (2015, background): “Along this direction, strong efforts have been spent recently by the international scientiﬁc community, through the development of corpora and challenges, such as REVERB [15], CHIME [16, 17] and ASpIRE.”
Can we Automatically Transform Speech Recorded on Common Consumer Devices in Real-World Environments into Professional Production Quality Speech?—A Dataset, Insights, and Challenges (2015, background): “A speech enhancement algorithm would directly [11] or indirectly [8]–[10] apply some form of decoloration and perhaps bandwidth expansion [12], [13].”
Speech acoustic modeling from raw multichannel waveforms (2015, background): “Results from the recent REVERB [6] and CHIME [7] challenges demonstrate how these techniques can be used in combination, e.g. [8] who combine a speech enhancement frontend comprised of linear prediction derever-beration, beamforming, and model-based noise reduction at the input to an ASR system.”
SpeakerBeam: Speaker Aware Neural Network for Target Speaker Extraction in Speech Mixtures (2019, neutral): “with the REVERB challenge [57].”
Batch-normalized joint training for DNN-based distant speech recognition (2016, background): “ recently, DIRHA which addressed speech-based domestic control. The progress in the ﬁeld was also fostered by the considerable success of some international challenges such as CHiME [5, 6] and REVERB [7]. Despite the great progress made in the past years, current systems still exhibit a signiﬁcant l”
The Automatic Speech recogition In Reverberant Environments (ASpIRE) challenge (2015, methodology): “The REVERB (REverberant Voice Enhancement and Recognition Benchmark) challenge [15] provided an evaluation framework with common datasets, tasks, and evaluation metrics for both speech enhancement and STT for speech recorded from stationary distant-talking speakers with 1-channel, 2-channel, or 8-ch”
Spatial diffuseness features for DNN-based speech recognition in noisy and reverberant environments (2014, methodology): “Finally, we evaluate the proposed feature on the two-channel task of the REVERB challenge [14], showing that the proposed approach outperforms both noisy multi-condition training and multichannel spectral subtraction-based signal enhancement.”
A network of deep neural networks for Distant Speech Recognition (2017, background): “The progress in the ﬁeld was also fostered by the considerable success of some international challenges such as CHiME [3] and REVERB [4].”
Reverberant speech recognition combining deep neural networks and deep autoencoders augmented with a phone-class feature (2015, methodology): “The proposed system was evaluated following the instructions for the ASR task of the Reverb Challenge 2014 [32].”
Deep recurrent de-noising auto-encoder and blind de-reverberation for reverberated speech recognition (2014, methodology): “Our method is evaluated on the 2014 REVERB Challenge data [14] which features both simulated reverberated and noisy data as well as real recordings from a meeting room.”
The MERL/MELCO/TUM system for the REVERB Challenge using Deep Recurrent Neural Network Feature Enhancement (2014, methodology): “Our methods are evaluated on the official development and evaluation sets of the 2014 REVERB Challenge [1].”
A study on data augmentation of reverberant speech for robust speech recognition (2017, methodology): “The set of real RIRs (R) is composed of three databases: the RWCP sound scene database [8], the REVERB challenge database [9] and the Aachen impulse response database [10].”
NARA-WPE: A Python package for weighted prediction error dereverberation in Numpy and Tensorflow for online and offline processing (2018, methodology): “To evaluate the effectiveness of the dereverberation implementation, we demonstrate ASR results on the REVERB challenge data [8].”
JHU ASpIRE system: Robust LVCSR with TDNNS, iVector adaptation and RNN-LMS (2015, methodology): “Multi-condition training data was created by distorting speech with real world room impulse response (RIR) and noise recordings available from three different databases viz., the RWCP sound scene database 3 [15], the REVERB challenge database [16] and the Aachen impulse response database [17].”
Student-teacher network learning with enhanced features (2017, background): “Speech enhancement and ASR techniques have been developing rapidly with the growing demand for distant-talking speech interfaces and recent challenge activities [1, 2, 3, 4].”
Improved MVDR Beamforming Using Single-Channel Mask Prediction Networks (2016, background): “Recent work on multi-microphone databases such as the AMI database [1, 2, 3] and recent challenges such as REVERB [4], CHiME-2 [5], and CHiME-3 [6] indicate that performing beamforming of multiple microphone data improves performance as compared to other direct techniques for speech recognition.”