The Diverse Environments Multi-channel Acoustic Noise Database (DEMAND): A database of multichannel environmental noise recordings

By Joachim Thiemann, Nobutaka Ito, Emmanuel Vincent. 2013

783 citations recorded by the source corpus.

Subfields: Biomedical Engineering, 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 Biomedical Engineering

In Signal Processing

Citations in the corpus

A Wavenet for Speech Denoising (2017, methodology): “The used dataset [22, 29] was generated from two sources: speech data was supplied by the Voice Bank corpus [32] while environmental sounds were provided by the Diverse Environments Multichannel Acoustic Noise Database (DEMAND) [28].”
SEGAN: Speech Enhancement Generative Adversarial Network (2017, background): “ included in the train set and 2 in the test set. To make the noisy training set, a total of 40 different conditions are considered [27]: 10 types of noise (2 artiﬁcial and 8 from the Demand database [29]) with 4 signal-to-noise ratio (SNR) each (15, 10, 5, and 0dB). There are around 10 different se”
Icassp 2022 Deep Noise Suppression Challenge (2022, methodology): “We augmented chosen Audioset noise clips with 10,000 noise clips downloaded from Freesound and DEMAND databases [12].”
ICASSP 2021 Deep Noise Suppression Challenge (2020, methodology): “We also augmented an additional 10,000 noise clips downloaded from Freesound and DEMAND databases [20].”
Deep Residual-Dense Lattice Network for Speech Enhancement (2020, methodology): “A total of 20 different conditions were used to create the noisy speech, including five noise types from the DEMAND dataset (Thiemann, Ito, and Vincent 2013), and four SNR levels: 2.5, 7.5, 12.5, and 17.5 dB.”
Speech Enhancement for a Noise-Robust Text-to-Speech Synthesis System Using Deep Recurrent Neural Networks (2016, methodology): “To create the noisy database used for training we used ten different types of noise: two artificially generated (speechshaped noise and babble) and eight real noise recordings from the Demand database [24].”
Speech Enhancement of Noisy and Reverberant Speech for Text-to-Speech (2018, methodology): “derived from the multichannel Demand database [26], more specifically from the first channel of the 48 kHz versions of”
Interspeech 2021 Deep Noise Suppression Challenge (2021, methodology): “We also augmented an additional 10,000 noise clips downloaded from Freesound and DEMAND databases [21].”
The INTERSPEECH 2020 Deep Noise Suppression Challenge: Datasets, Subjective Testing Framework, and Challenge Results (2020, methodology): “10,000 noise clips downloaded from Freesound and DEMAND databases [19].”
Investigating RNN-based speech enhancement methods for noise-robust Text-to-Speech (2016, positive): “To create the noisy database used for training we used ten different noise types: two artificially generated (speech-shaped noise and babble) and eight real noise recordings from the Demand database [25].”
PHASEN: A Phase-and-Harmonics-Aware Speech Enhancement Network (2019, methodology): “Voice Bank+DEMAND: This is an open dataset 1 proposed by (Valentini-Botinhao et al. 2016).”
ICASSP 2022 Acoustic Echo Cancellation Challenge (2022, methodology): “The noisy speech dataset was created by mixing clean speech with noise clips sampled from Audioset [16], Freesound 2 and DEMAND [17] databases at signal to noise ratios sampled uniformly from [0, 40] dB.”
Time-Frequency Masking-Based Speech Enhancement Using Generative Adversarial Network (2018, background): “The total of 40 different noisy conditions with 10 types of noise (2 artificial and 8 from Demand database [22]) and 4 signal-to-noise ratio (SNR) each (15, 10, 5, and 0 dB) are considered for the noisy training set.”
Speech Enhancement and Dereverberation With Diffusion-Based Generative Models (2022, neutral): “The utterances are artificially contaminated with eight real-recorded noise samples from the DEMAND database [54] and two artificially generated noise samples (babble and speech shaped) at SNRs of 0, 5, 10, and 15 dB.”
Improved Speech Enhancement with the Wave-U-Net (2018, positive): “These are composed of 10 types of noise (2 of which are artiﬁcially-generated 1 and 8 sourced from the DEMAND database [13], each mixed with clean speech at one of 4 signal-to-noise ratios (SNR) (15, 10, 5, and 0 dB).”
The INTERSPEECH 2020 Deep Noise Suppression Challenge: Datasets, Subjective Speech Quality and Testing Framework (2020, methodology): “We also augmented an additional 10,000 noise clips downloaded from Freesound and DEMAND databases [17].”
NISQA: A Deep CNN-Self-Attention Model for Multidimensional Speech Quality Prediction with Crowdsourced Datasets (2021, methodology): “To simulate real background noises, the noise clips from the DNS-Challenge datasets were used, which in turn are taken from the Audioset [28], freesound [29], and DEMAND [30] corpora.”