Pith. sign in

REVIEW 4 major objections 3 minor 41 references

Adaptive Data Augmentation with NaturalSpeech3 for Far-field Speaker Verification

T0 review · 4 major / 3 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Replacing one embedding in a pretrained codec converts near-field speech into far-field training data, cutting FFSVC2020 equal error rates to as low as 5.7 percent.

desk verdict A neat codec-based augmentation idea with a valid train-only result, but the headline test-time result rests on a trial-dependent enrollment protocol and should not be taken at face value. read the letter →

arxiv 2501.08691 v1 pith:MCIZ3TFY submitted 2025-01-15 cs.SD eess.AS

Li Zhang , Jiyao Liu , Lei Xie This is my paper · ORCID
classification cs.SDeess.AS
keywords far-fieldspeakerverificationdataaugmentationNaturalSpeech3FACodecvoiceconversionECAPA-TDNNFFSVC2020embeddingdisentanglement
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that a pretrained speech-generation codec can serve as a data augmentation engine for far-field speaker verification: by decomposing near- and far-field speech into content, prosody, speaker, and residual acoustic embeddings, and replacing only the speaker embedding of far-field speech with one from near-field speech, the reconstructed audio carries the near-field speaker's identity inside a far-field acoustic environment. On the FFSVC2020 benchmark this pseudo far-field speech lowers equal error rates from 10.00%, 14.06%, and 8.711% to 6.022%, 7.669%, and 5.707% when the same swap is applied to enrollment and test trials, beating random noise, reverberation, SpecAugment, filter augmentation, and a StarGAN-based voice-conversion augmentation. The practical stake is that speaker-annotated far-field data is scarce, so a method that converts abundant near-field speech into environment-matched far-field speech could remove a bottleneck in training robust far-field speaker verification systems.

What carries the argument

The load-bearing object is FACodec, the factorized vector-quantization codec from NaturalSpeech3, which decomposes a waveform into four embedding subspaces: content, prosody, speaker, and residual acoustic details. The method's operation is one embedding swap: keep $F^f_p$, $F^f_c$, and $F^f_r$ from the far-field utterance, replace $F^f_s$ with $F^n_s$ from a near-field utterance, and reconstruct with the NaturalSpeech3 voice-conversion module. The residual subspace is what is supposed to carry room acoustics and environmental noise; the speaker subspace is what is supposed to carry identity; the swap transfers identity without transferring the room.

What would settle it

Take a single far-field utterance, swap in speaker embeddings from several near-field speakers, and reconstruct; then run a speaker identification trial and a blind RT60 estimate on the outputs. If the reconstructed speech is not attributed to the near-field donor above chance, or if its RT60 no longer matches the far-field source room, the embedding swap is not doing the causal work the paper attributes to it.

Watch

Extended reading notes

Core claim

The paper's central claim is that the four-way factorization learned by FACodec is clean enough that a single embedding substitution implements a controllable voice-and-environment transfer: with far-field speech as the acoustic carrier and near-field speech as the identity donor, $S_{\text{pseudo-far}} = C(F^f_p, F^f_c, F^n_s, F^f_r)$ yields speech that a speaker encoder attributes to the near-field speaker while a blind RT60 estimator places it in the far-field room. The same mechanism is used two ways: to enlarge the training set by transferring all AISHELL-2 speaker identities onto FFSVC2020 utterances, and to augment enrollment and test trials so that enrollment and test speech share a converted common condition. In both uses the paper reports that this adaptive augmentation achieves lower EER and minDCF than in-domain noise and reverberation, out-domain VoxCeleb data, SpecAugment, speed perturbation, shuffle augmentation, filter augmentation, and VC2Aug on all three FFSVC2020 tasks.

Load-bearing premise

The whole method rests on the assumption that FACodec's speaker embedding contains only the speaker's identity and that the content, prosody, and residual embeddings contain all of the text and the room acoustics, so swapping in a new speaker vector changes nothing else.

Editorial extensions

If this is right

  • If the swap works as claimed, any large near-field corpus with speaker labels can be repurposed as far-field training data, multiplying the number of speaker identities in a far-field SV system from $q$ to $q+t$.
  • Applying the same conversion to enrollment and test utterances should reduce trial mismatch caused by text and environment differences, which is the mechanism behind the reported test-time gains.
  • The pseudo far-field speech should inherit the target room's reverberation characteristics rather than the donor speaker's, as the RT60 analysis indicates, so the method should generalize to new far-field deployment rooms without manual SNR or RIR tuning.
  • Because the method builds on a pretrained generative codec, it can be applied without retraining or fine-tuning the codec for each new far-field dataset, and it adds new speaker identities in a way that the paper argues StarGAN-based VC2Aug does not.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: the clean-factorization assumption is directly testable; if a trained speaker encoder can reliably pick the near-field donor out of a line-up from the pseudo speech while an environment classifier attributes the pseudo speech to the far-field room, the swap mechanism is confirmed, and if either fails, the reported gains may come from extra data volume or from the reconstruction itself r
  • Inference: the same embedding-swap recipe could be ported to other codec-based factorization models or to far-field automatic speech recognition and keyword spotting, where domain-matched audio is also scarce.
  • Inference: a stress test that swaps speaker embeddings across far-field utterances recorded in very different rooms, such as a small office versus a hall, would isolate whether the residual embeddings actually encode the room rather than merely some global channel effect.
  • Inference: the reported test-time variant implies that converting enrollment and test speech into a common pseudo-domain improves score comparability, which could be probed independently by measuring score distributions before and after conversion.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 3 minor

Summary. The paper proposes an adaptive data augmentation method for far-field speaker verification. It uses NaturalSpeech3's FACodec to decompose near-field and far-field speech into content, prosody, speaker, and residual embeddings, then combines non-speaker embeddings from far-field speech with speaker embeddings from near-field speech to synthesize pseudo far-field speech. The method is evaluated on FFSVC2020 with an ECAPA-TDNN backbone, reporting lower EER and minDCF than several augmentation baselines. The paper presents two variants: a training-time augmentation and a test-time augmentation that also transforms enrollment and test utterances. The central claim is that the proposed adaptive augmentation significantly outperforms traditional data augmentation strategies.

Significance. If the results were obtained under a valid protocol, the paper would offer a timely and useful way to leverage near-field speech for far-field speaker verification by exploiting a pretrained factorized codec. The training-time results in Table I are consistently better than the listed baselines, and the RT60 analysis in Section V.B is a reasonable sanity check for acoustic-environment preservation. However, the headline test-time result is produced by a protocol that leaks trial-specific information into the enrollment, and no statistical significance testing is reported. The paper's central claim therefore rests on an invalid evaluation and is not currently supported.

major comments (4)
  1. [Section V.A, Table I] The row 'Adaptive Data Augmentation (Test)' is obtained from a trial-dependent protocol: Section V.A states that 'the speaker identity from the enrollment set is converted into the speaker identity of the test set (and vice versa), generating augmented samples for testing.' This makes the enrollment representation depend on the test utterance of each trial, violating the fixed enrollment/test split required in speaker verification evaluation. The reported EER values (6.022%, 7.669%, 5.707%) are therefore not a valid comparison against the baselines, and the abstract's claim of 'significantly outperforms' relies on this invalid row.
  2. [Section IV.D and Table I] No standard deviations, confidence intervals, or statistical significance tests are reported for any result. Since the differences between the proposed method and the strongest baseline are small (e.g., 6.250% vs. 6.990% on Task1 for the training variant), a single run is insufficient to support the word 'significantly' in the abstract. The paper should report means over multiple random seeds or a proper significance test for the key comparisons.
  3. [Section III.B, Eq. (3)] The method assumes that FACodec's speaker embedding F_s contains only speaker identity and that the prosody, content, and residual embeddings F_p, F_c, F_r preserve the far-field environment. This disentanglement assumption is load-bearing, but the paper provides no direct verification that the pseudo far-field speech preserves the near-field speaker's identity or the far-field environment. I recommend adding an experiment that verifies the converted utterances against the original near-field speaker and an ablation study showing the contribution of each embedding component.
  4. [Section V.B] The RT60 analysis is only qualitative and does not establish that the adaptive augmentation is superior to simpler environment-aware baselines. For example, the method should be compared against data augmentation that uses measured RIRs from the far-field recording room or noise estimated from the far-field utterances. Without such a comparison, the claimed advantage of 'adaptivity' over traditional augmentation is not demonstrated.
minor comments (3)
  1. [Abstract and Section I] The phrase 'difficult to simulate, and can significantly affect' is redundant, and the sentence 'the collection of far-field speech is more expensive and time-consuming compared with near-field speech collection' is wordy and should be revised.
  2. [Section III.B] Equation (3) is unnumbered and contains a typo ('Spesudo-far' should be 'S_pseudo-far'). Please number the equation and correct the typo.
  3. [Section III.C] The notation for pseudo far-field speaker labels is inconsistent: it shows 'y^s_3' in the set definition, which should be 'y^n_3' to match the other elements.

Circularity Check

2 steps flagged · score 6.0 of 10

Part of the evidence is self-referential: test-time trials are built from the test speaker identity, and the RT60 check reuses the copied residual; the training-row result remains independent.

  1. self definitional [Section V.A, Table I, 'Adaptive Data Augmentation (Test)' row]
    "In this method, the speaker identity from the enrollment set is converted into the speaker identity of the test set (and vice versa), generating augmented samples for testing."

    The enrollment side of each trial is constructed from the test side's speaker identity. The EER reported for the last Table I row is therefore not an independent verification result: the reference enrollment already contains the identity information of the utterance it is compared against. The improvement to 6.022%/7.669%/5.707% is obtained by injecting test-side identity into the enrollment, so the 'prediction' is partly defined by the input it is supposed to verify.

  2. self definitional [Section V.B, Augmented Sample Adaptive Analysis, Fig. 3]
    "As shown in Fig. 3, the RT60 values of the augmented speech align more closely with those of the original far-field audio, even when the speech is converted from far-field to near-field."

    The pseudo far-field speech is constructed as S_pseudo-far = C(F_p^f, F_c^f, F_s^n, F_r^f), so it reuses the far-field residual embedding F_r^f, the component designated to carry acoustic details. RT60 is a room-acoustic property carried by that residual. Matching RT60 between the pseudo and real far-field speech is therefore essentially guaranteed by the construction, not evidence that the model has independently 'adapted' to the acoustic environment. The validation metric is the same quantity copied into the generated sample.

full rationale

The training-time augmentation core is not circular: it uses NaturalSpeech3's external FACodec to combine near-field speaker embeddings with far-field non-speaker embeddings, and the 'Adaptive Data Augmentation (Train)' row is evaluated on fixed FFSVC2020 trials without using test identities. Self-citations such as [10], [18], and [21] are contextual and not load-bearing; no fitted parameter is renamed as a prediction. However, the abstract and Section V explicitly promote the test-time variant as part of the claimed superiority, and that variant defines the enrollment using the test utterance's speaker identity, making the trial self-referential. The RT60 'adaptability' check is also self-confirming because the pseudo far-field waveform reuses the far-field residual embedding by construction. These two steps make part of the evidence circular, so the overall circularity score is 6, while the independent training-augmentation claim keeps the score below 8.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The method does not introduce new fitted constants or new physical entities. It borrows the pretrained NaturalSpeech3/FACodec factorization and the ECAPA-TDNN backbone, and its central assumption is that the factorization is truly disentangled. The number of pseudo speakers selected from AISHELL-2 is an unstated experimental choice, and the validity of the test-time protocol is assumed without evidence.

free parameters (1)
  • Number of selected pseudo far-field speakers from AISHELL-2 (denoted t) = unspecified; AISHELL-2 has 1991 speakers
    Section III-C states that the method shuffles and selects data corresponding to t speakers from the q by t pool. The exact value of t and the utterance-level pairing/selection policy are not stated, yet they determine the composition of the augmented training set and therefore the reported EER.
assumptions (4)
  • domain assumption FACodec's learned subspaces are disentangled: F_s captures speaker identity and not environment, content, or prosody.
    Invoked in Section III-B, where C(F_p^f, F_c^f, F_s^n, F_r^f) is assumed to preserve far-field environment and transfer near-field identity.
  • domain assumption The residual embedding F_r is sufficient to reproduce the far-field acoustic environment after reconstruction.
    Section III-B and the RT60 analysis in Section V-B rely on residual acoustics being preserved when only F_s is swapped.
  • ad hoc to paper The test-time augmentation protocol is a valid speaker verification evaluation procedure.
    Section V-A uses enrollment/test augmentation from the trial pair itself; no justification is given that this matches deployment conditions.
  • domain assumption Synthetic pseudo far-field speech contains no artifacts that let the speaker encoder exploit generation fingerprints.
    No objective artifact or speaker-similarity evaluation of the pseudo audio is reported in Section V.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Adaptive Data Augmentation with NaturalSpeech3 for Far-field Speaker Verification." pith.science (2026). https://pith.science/paper/MCIZ3TFY

@misc{pith2026250108691,
  author       = {Pith},
  title        = {Pith review of: Adaptive Data Augmentation with NaturalSpeech3 for Far-field Speaker Verification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MCIZ3TFY}},
  note         = {Machine review of arXiv:2501.08691}
}
read the original abstract

The scarcity of speaker-annotated far-field speech presents a significant challenge in developing high-performance far-field speaker verification (SV) systems. While data augmentation using large-scale near-field speech has been a common strategy to address this limitation, the mismatch in acoustic environments between near-field and far-field speech significantly hinders the improvement of far-field SV effectiveness. In this paper, we propose an adaptive speech augmentation approach leveraging NaturalSpeech3, a pre-trained foundation text-to-speech (TTS) model, to convert near-field speech into far-field speech by incorporating far-field acoustic ambient noise for data augmentation. Specifically, we utilize FACodec from NaturalSpeech3 to decompose the speech waveform into distinct embedding subspaces-content, prosody, speaker, and residual (acoustic details) embeddings-and reconstruct the speech waveform from these disentangled representations. In our method, the prosody, content, and residual embeddings of far-field speech are combined with speaker embeddings from near-field speech to generate augmented pseudo far-field speech that maintains the speaker identity from the out-domain near-field speech while preserving the acoustic environment of the in-domain far-field speech. This approach not only serves as an effective strategy for augmenting training data for far-field speaker verification but also extends to cross-data augmentation for enrollment and test speech in evaluation trials.Experimental results on FFSVC demonstrate that the adaptive data augmentation method significantly outperforms traditional approaches, such as random noise addition and reverberation, as well as other competitive data augmentation strategies.

Figures

Figures reproduced from arXiv: 2501.08691 by the authors.

Figure 1
Figure 1. The Overview of the Adaptive Data Augmentation. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. EER Comparison Across FFSVC2020 Tasks for Different Augmen [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Visualization of Estimated RT60 for Real Speech and Adaptively Augmented Speech. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 33 canonical work pages

  1. [1]

    An overview of automatic speaker verification system

    Ravika Naika. An overview of automatic speaker verification system. In Intelligent Computing and Information and Communication: Pro- ceedings of 2nd International Conference, ICICC 2017 , pages 603–610. Springer, 2018

    work page 2017

  2. [2]

    V oxceleb: Large-scale speaker verification in the wild.Computer Speech & Language, 60:101027, 2020

    Arsha Nagrani, Joon Son Chung, Weidi Xie, and Andrew Zisserman. V oxceleb: Large-scale speaker verification in the wild.Computer Speech & Language, 60:101027, 2020

    2020

  3. [3]

    Aishell-1: An open-source mandarin speech corpus and a speech recognition baseline

    Hui Bu, Jiayu Du, Xingyu Na, Bengu Wu, and Hao Zheng. Aishell-1: An open-source mandarin speech corpus and a speech recognition baseline. In 2017 20th conference of the oriental chapter of the international coordinating committee on speech databases and speech I/O systems and assessment (O-COCOSDA) , pages 1–5. IEEE, 2017

    work page 2017

  4. [4]

    Aishell-2: Trans- forming mandarin asr research into industrial scale

    Jiayu Du, Xingyu Na, Xuechen Liu, and Hui Bu. Aishell-2: Trans- forming mandarin asr research into industrial scale. arXiv preprint arXiv:1808.10583, 2018

    arXiv 2018

  5. [5]

    Intel far-field speaker recognition system for voices challenge 2019

    Jonathan Huang and Tobias Bocklet. Intel far-field speaker recognition system for voices challenge 2019. In Interspeech, pages 2473–2477, 2019

    work page 2019

  6. [6]

    Far-field speaker verification challenge (ffsvc) 2022: Challenge evaluation plan, 2022

    Xiaoyi Qin, Ming Li, Hui Bu, Shrikanth Narayanan, and Haizhou Li. Far-field speaker verification challenge (ffsvc) 2022: Challenge evaluation plan, 2022

    work page 2022

  7. [7]

    The voices from a distance challenge 2019

    Mahesh Kumar Nandwana, Julien Van Hout, Colleen Richey, Mitchell McLaren, Maria Alejandra Barrios, and Aaron Lawson. The voices from a distance challenge 2019. In Interspeech, pages 2438–2442, 2019

    work page 2019

  8. [8]

    Hi-mia: A far-field text-dependent speaker verification database and the baselines

    Xiaoyi Qin, Hui Bu, and Ming Li. Hi-mia: A far-field text-dependent speaker verification database and the baselines. In ICASSP 2020-2020, pages 7609–7613. IEEE, 2020

    work page 2020

Show all 41 references
  1. [9]

    Multisv: Dataset for far-field multi-channel speaker verifi- cation

    Ladislav Mo ˇsner, Old ˇrich Plchot, Luk ´aˇs Burget, and Jan Honza ˇCernock`y. Multisv: Dataset for far-field multi-channel speaker verifi- cation. In ICASSP 2022-2022, pages 7977–7981. IEEE, 2022

    2022

  2. [10]

    Npu speaker verification system for interspeech 2020 far-field speaker verification challenge

    Li Zhang, Jian Wu, and Lei Xie. Npu speaker verification system for interspeech 2020 far-field speaker verification challenge. arXiv preprint arXiv:2008.03521, 2020

    2020 arXiv

  3. [11]

    The interspeech 2020 far-field speaker verification challenge

    Xiaoyi Qin, Ming Li, Hui Bu, Wei Rao, Rohan Kumar Das, Shrikanth Narayanan, and Haizhou Li. The interspeech 2020 far-field speaker verification challenge. arXiv preprint arXiv:2005.08046 , 2020

    2020 arXiv

  4. [12]

    Multi-channel speaker verification with conv-tasnet based beamformer

    Ladislav Mo ˇsner, Old ˇrich Plchot, Luk ´aˇs Burget, and Jan Honza ˇCernock`y. Multi-channel speaker verification with conv-tasnet based beamformer. In ICASSP 2022-2022, pages 7982–7986. IEEE, 2022

    2022

  5. [13]

    Multi- channel far-field speaker verification with large-scale ad-hoc microphone arrays

    Chengdong Liang, Yijiang Chen, Jiadi Yao, and Xiao-Lei Zhang. Multi- channel far-field speaker verification with large-scale ad-hoc microphone arrays. arXiv preprint arXiv:2110.05975 , 2021

    2021 arXiv

  6. [14]

    Unet-densenet for robust far-field speaker verification

    Zhenke Gao, Man-Wai Mak, and Weiwei Lin. Unet-densenet for robust far-field speaker verification. In Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH , volume 2022, pages 3714–3718, 2022

    2022

  7. [15]

    Dereverberation and beamforming in robust far-field speaker recognition

    Ladislav Mo ˇsner, Oldrich Plchot, Pavel Matejka, Ondrej Novotn `y, and Jan Cernock `y. Dereverberation and beamforming in robust far-field speaker recognition. In Proc. Interspeech, volume 2018, pages 1334– 1338, 2018

    2018

  8. [16]

    Far-field end-to-end text- dependent speaker verification based on mixed training data with transfer learning and enrollment data augmentation

    Xiaoyi Qin, Danwei Cai, and Ming Li. Far-field end-to-end text- dependent speaker verification based on mixed training data with transfer learning and enrollment data augmentation. In Interspeech, pages 4045– 4049, 2019

    2019

  9. [17]

    Stc-innovation speaker recognition systems for far- field speaker verification challenge 2020

    Aleksei Gusev, Vladimir V olokhov, Alisa Vinogradova, Tseren Andzhukaev, Andrey Shulipa, Sergey Novoselov, Timur Pekhovsky, and Alexander Kozlov. Stc-innovation speaker recognition systems for far- field speaker verification challenge 2020. In INTERSPEECH, pages 3466–3470, 2020

    2020

  10. [18]

    Npu-hc speaker verification system for far-field speaker verification challenge 2022

    Li Zhang, Yue Li, Namin Wang, Jie Liu, and Lei Xie. Npu-hc speaker verification system for far-field speaker verification challenge 2022. INTERSPEECH, 2022

    2022

  11. [19]

    Specaugment: A simple data augmentation method for automatic speech recognition

    Daniel S Park, William Chan, Yu Zhang, Chung-Cheng Chiu, Barret Zoph, Ekin D Cubuk, and Quoc V Le. Specaugment: A simple data augmentation method for automatic speech recognition. arXiv preprint arXiv:1904.08779, 2019

    1904 arXiv

  12. [20]

    Robust speaker verification using population-based data augmentation

    Weiwei Lin and Man-Wai Mak. Robust speaker verification using population-based data augmentation. In ICASSP 2022-2022 , pages 7642–7646. IEEE, 2022

    2022

  13. [21]

    Distance-based weight transfer for fine-tuning from near- field to far-field speaker verification

    Li Zhang, Qing Wang, Hongji Wang, Yue Li, Wei Rao, Yannan Wang, and Lei Xie. Distance-based weight transfer for fine-tuning from near- field to far-field speaker verification. In ICASSP 2023-2023, pages 1–5. IEEE, 2023

    2023

  14. [22]

    Naturalspeech 3: Zero-shot speech synthesis with factorized codec and diffusion models

    Zeqian Ju, Yuancheng Wang, Kai Shen, Xu Tan, Detai Xin, Dongchao Yang, Yanqing Liu, Yichong Leng, Kaitao Song, Siliang Tang, et al. Naturalspeech 3: Zero-shot speech synthesis with factorized codec and diffusion models. arXiv preprint arXiv:2403.03100 , 2024

    2024 arXiv

  15. [23]

    X-vectors: Robust dnn embeddings for speaker recognition

    David Snyder, Daniel Garcia-Romero, Gregory Sell, Daniel Povey, and Sanjeev Khudanpur. X-vectors: Robust dnn embeddings for speaker recognition. In ICASSP 2018, pages 5329–5333. IEEE, 2018

    2018

  16. [24]

    Specaugment impact on automatic speaker verification system

    Muhammad Yusuf Faisal and Suyanto Suyanto. Specaugment impact on automatic speaker verification system. In 2019 International Seminar on Research of Information Technology and Intelligent Systems (ISRITI), pages 305–308. IEEE, 2019

    2019

  17. [25]

    Investigation of data augmen- tation techniques for disordered speech recognition

    Mengzhe Geng, Xurong Xie, Shansong Liu, Jianwei Yu, Shoukang Hu, Xunying Liu, and Helen Meng. Investigation of data augmen- tation techniques for disordered speech recognition. arXiv preprint arXiv:2201.05562, 2022

    2022 arXiv

  18. [26]

    Improved far-field speaker recognition method based geometry acoustic simulation and specaugment

    Yingzi Lian and Jing Pang. Improved far-field speaker recognition method based geometry acoustic simulation and specaugment. In 2021 International Conference on Intelligent Computing, Automation and Applications (ICAA), pages 380–387. IEEE, 2021

    2021

  19. [27]

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

    Tom Ko, Vijayaditya Peddinti, Daniel Povey, Michael L Seltzer, and Sanjeev Khudanpur. A study on data augmentation of reverberant speech for robust speech recognition. In ICASSP 2017, pages 5220–5224. IEEE, 2017

    2017

  20. [28]

    Synaug: Synthesis-based data augmentation for text-dependent speaker verification

    Chenpeng Du, Bing Han, Shuai Wang, Yanmin Qian, and Kai Yu. Synaug: Synthesis-based data augmentation for text-dependent speaker verification. In ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages 5844–5848. IEEE, 2021

    2021

  21. [29]

    Overcoming data scarcity in speaker identification: Dataset augmentation with synthetic mfccs via character-level rnn

    Jordan J Bird, Diego R Faria, Cristiano Premebida, Anik ´o Ek ´art, and Pedro PS Ayrosa. Overcoming data scarcity in speaker identification: Dataset augmentation with synthetic mfccs via character-level rnn. In 2020 IEEE International Conference on Autonomous Robot Systems and...

    2020

  22. [30]

    Data augmentation using deep generative models for embedding based speaker recognition

    Shuai Wang, Yexin Yang, Zhanghao Wu, Yanmin Qian, and Kai Yu. Data augmentation using deep generative models for embedding based speaker recognition. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 28:2598–2609, 2020

    2020

  23. [31]

    Data Aug- mentation Using Variational Autoencoder for Embedding Based Speaker Verification

    Zhanghao Wu, Shuai Wang, Yanmin Qian, and Kai Yu. Data Aug- mentation Using Variational Autoencoder for Embedding Based Speaker Verification. In Proc. Interspeech 2019 , pages 1163–1167, 2019

    2019

  24. [32]

    Stargan-vc based cross- domain data augmentation for speaker verification

    Hang-Rui Hu, Yan Song, Jian-Tao Zhang, Li-Rong Dai, Ian McLoughlin, Zhu Zhuo, Yu Zhou, Yu-Hong Li, and Hui Xue. Stargan-vc based cross- domain data augmentation for speaker verification. In ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Proces...

    2023

  25. [33]

    Ecapa- tdnn: Emphasized channel attention, propagation and aggregation in tdnn based speaker verification

    Brecht Desplanques, Jenthe Thienpondt, and Kris Demuynck. Ecapa- tdnn: Emphasized channel attention, propagation and aggregation in tdnn based speaker verification. arXiv preprint arXiv:2005.07143 , 2020

    2005 arXiv

  26. [34]

    Arcface: Additive angular margin loss for deep face recognition

    Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou. Arcface: Additive angular margin loss for deep face recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4690–4699, 2019

    2019

  27. [35]

    Musan: A music, speech, and noise corpus

    David Snyder, Guoguo Chen, and Daniel Povey. Musan: A music, speech, and noise corpus. arXiv preprint arXiv:1510.08484 , 2015

    2015 arXiv

  28. [36]

    Room impulse response generator

    Emanuel AP Habets. Room impulse response generator. Technische Universiteit Eindhoven, Tech. Rep , 2(2.4):1, 2006

    2006

  29. [37]

    Speaker augmentation and bandwidth extension for deep speaker embedding

    Hitoshi Yamamoto, Kong Aik Lee, Koji Okabe, and Takafumi Koshi- naka. Speaker augmentation and bandwidth extension for deep speaker embedding. In Interspeech, pages 406–410, 2019

    2019

  30. [38]

    Shuffleaug- ment: A data augmentation method using time shuffling

    Yoshinao Sato, Narumitsu Ikeda, and Hirokazu Takahashi. Shuffleaug- ment: A data augmentation method using time shuffling. In ICASSP 2023-2023, pages 1–5. IEEE, 2023

    2023

  31. [39]

    Filteraugment: An acoustic environmental data augmentation method

    Hyeonuk Nam, Seong-Hu Kim, and Yong-Hwa Park. Filteraugment: An acoustic environmental data augmentation method. In ICASSP 2022- 2022, pages 4308–4312. IEEE, 2022

    2022

  32. [40]

    Stargan-vc based cross- domain data augmentation for speaker verification

    Hang-Rui Hu, Yan Song, Jian-Tao Zhang, Li-Rong Dai, Ian McLoughlin, Zhu Zhuo, Yu Zhou, Yu-Hong Li, and Hui Xue. Stargan-vc based cross- domain data augmentation for speaker verification. In ICASSP 2023- 2023 IEEE International Conference on Acoustics, Speech and Signal Process...

    2023

  33. [41]

    Blind estimation of reverberation time

    Rama Ratnam, Douglas L Jones, Bruce C Wheeler, William D O’Brien Jr, Charissa R Lansing, and Albert S Feng. Blind estimation of reverberation time. The Journal of the Acoustical Society of America , 114(5):2877–2892, 2003

    2003

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.