A study on data augmentation of reverberant speech for robust speech recognition

By Tom Ko, Vijayaditya Peddinti, Daniel Povey, M. Seltzer, Sanjeev Khudanpur. 2017

1,178 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Speech & Audio

In Signal Processing

Citations in the corpus

FLEURS: FEW-Shot Learning Evaluation of Universal Representations of Speech (2022, background): “SpecAugment [23], speed perturbation [24], simulated reverberation [25], etc.”
Fairseq S2T: Fast Speech-to-Text Modeling with Fairseq (2020, methodology): “For further pre-processing, FAIRSEQ S2T provides online speech data transforms, including CMVN (cepstral mean and variance normalization), speed perturbation (Ko et al., 2017) and SpecAugment (Park et al., 2019).”
Privacy-Preserving Adversarial Representation Learning in ASR: Reality or Illusion? (2019, positive): “The train-960 subset was augmented using room impulse responses, isotropic and point-source noises [34] as well as music and speech [35] as per the standard sre16 recipe for training x-vectors [24] from the Kaldi toolkit [36], which we adapted to Librispeech.”
UniAudio: An Audio Foundation Model Toward Universal Audio Generation (2023, methodology): “For the text-to-music task, we follow MusicGen Copet et al. (2023), evaluating our methods on MusicCaps Agostinelli et al. (2023).”
WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing (2021, methodology): “Following the data simulation setup in [64], all the speech data from Switchboard-2 (Phase I & II & III), Switchboard Cellular (Part 1 & 2), the NIST Speaker Recognition Evaluation (2004 & 2005 & 2006 & 2008), the noises from [54] and the simulated room impulse responses used in [65] are leveraged f”
Icassp 2022 Deep Noise Suppression Challenge (2022, methodology): “We chose impulse responses from openSLR26 [13] 5 and openSLR28 [13] 6 datasets.”
Data Augmenting Contrastive Learning of Speech Representations in the Time Domain (2020, neutral): “Another one consists of adding noise, convolved with a room impulse response function to simulate point sources spread in space [16].”
Delving into VoxCeleb: environment invariant speaker recognition (2019, methodology): “pularised in computer vision [23, 24] by using label preserving transformations to augment image data. The method has been adopted for speaker recognition [18] by adding noise [25] and reverberations [26] to the speech signal. This recipe has been used by many of the recent work in the ﬁeld [27, 28,”
Lattice-free State-level Minimum Bayes Risk Training of Acoustic Models (2018, methodology): “The training data included 77 hours of meeting speech and was augmented six times using speed perturbation (x3)[21] and noise/reverberation perturbation (x2) [22].”
Is Someone Speaking?: Exploring Long-term Temporal Features for Audio-visual Active Speaker Detection (2021, positive): “As the Columbia ASD dataset is an open dataset, we apply the additional sources from RIRs data [25] and the MUSAN dataset [42] to perform audio augmentation on the TalkSet during training, and evaluate the performance using Sklearn library5.”
End-to-End Neural Speaker Diarization with Permutation-Free Objectives (2019, methodology): “The set of room impulse responses (RIRs) is the Simulated Room Impulse Response Database used in [34].”
The SpeakIn System for VoxCeleb Speaker Recognition Challange 2021 (2021, methodology): “Both foreground and background noises are added, and they are randomly selected from MUSAN and RIRs noises.”
An Investigation into the Effectiveness of Enhancement in ASR Training and Test for Chime-5 Dinner Party Transcription (2019, background): “[9] found that adding simulated reverberated speech improves accuracy significantly on several large vocabulary tasks.”
End-to-End Speaker Diarization for an Unknown Number of Speakers with Encoder-Decoder Based Attractors (2020, methodology): “For the training and evaluation, we used simulated mixtures created from Switchboard-2 (Phase I & II & III), Switchboard Cellular (Part 1 & 2), and the NIST Speaker Recognition Evaluation (2004 & 2005 & 2006 & 2008) for speech and the MUSAN corpus [28] for noise with simulated room impulse responses”
ASVspoof 2019: Future Horizons in Spoofed and Fake Audio Detection (2019, methodology): “Inspired by work to analyse and improve ASV reliability in re-verberant conditions [13, 14] and a similar approach used in the study of replay reported in [15], both bona ﬁde data and spoofed data contained in the PA database are generated according to a simulation [16, 17, 18] of their…”
SpeechMix - Augmenting Deep Sound Recognition Using Hidden Space Interpolations (2020, background): “Some of these include altering the shape or property [5, 6] and generating ex-ternal data for augmentation [7, 8, 9].”
X-Vectors: Robust DNN Embeddings for Speaker Recognition (2018, methodology): “in [25], and the reverberation itself is performed with the multicondition training tools in the Kaldi ASpIRE recipe [21].”
The Idlab Voxsrc-20 Submission: Large Margin Fine-Tuning and Quality-Aware Score Calibration in DNN Based Speaker Verification (2020, methodology): “We also create 6 additional augmented copies using the MUSAN [25] corpus (babble, noise), the RIR [26] (reverb) dataset and the SoX (tempo up, tempo down) and FFmpeg (compression) libraries.”
SLURP: A Spoken Language Understanding Resource Package (2020, background): “…among others, LibriSpeech (Panayotov et al., 2015), Switchboard (Godfrey et al., 1992), Fisher (Cieri et al., 2004), CommonVoice (Ardila et al., 2019), AMI (Carletta, 2007) and ICSI (Janin et al., 2003), which is further augmented to increase environmental robustness following (Ko et al., 2017) 7 .”