Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Efficient Speech Enhancement via Embeddings from Pre-trained Generative Audioencoders

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Denoising in the embedding space of a frozen generative audioencoder gives a small-network speech enhancer that beats a waveform-domain baseline in perceived quality.

desk verdict A clean, parameter-efficient embedding-denoiser recipe; the generative-vs-discriminative comparison is the real finding, but absolute gains rest on a vocoder whose reconstruction behavior isn't fully verified. read the letter →

arxiv 2506.11514 v1 pith:V5DZLA2A submitted 2025-06-13 eess.AS cs.SD

classification eess.AScs.SD
keywords speechenhancementaudioembeddingspre-trainedaudioencodervocodersynthesisspeakerfidelityembedding-spacedenoisingparameterefficiency
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 proposes treating speech enhancement as embedding-space denoising rather than waveform or time-frequency masking. A frozen pre-trained generative audioencoder maps noisy speech to embeddings; a compact denoise encoder (as small as two MLP layers) moves those embeddings toward clean ones; a pre-trained vocoder synthesizes the waveform. The authors claim this generative-embedding route preserves speaker and timbre information better than discriminative audioencoders, and that it surpasses the waveform-to-waveform Demucs model in subjective listening tests, with MOS 3.87 versus 3.11. The practical payoff is that neither the audioencoder nor the vocoder needs fine-tuning, leaving a lightweight, swappable module for restoration tasks.

What carries the argument

The load-bearing mechanism is the invertibility of the frozen generative audioencoder's embedding space: the Dasheng encoder outputs a 768-dimensional embedding from which a Vocos-style vocoder trained on clean speech can resynthesize audio, so signal enhancement reduces to regression in embedding space. The denoise encoder is trained with mean squared error between embeddings of paired noisy and clean speech, while the vocoder is trained separately on clean speech alone; a compact ViT-based network, and in the ablation even two MLP layers with 1.2M parameters, performs the regression. Keeping both the audioencoder and vocoder frozen is what gives the method its parameter efficiency and makes the middle module replaceable for other tasks.

What would settle it

A single experiment would settle the identity-preservation claim: take a set of unseen speakers, enhance their noisy speech with the full Dasheng pipeline, and run speaker verification between each clean utterance and its enhanced version; if equal-error rates are far worse than clean-versus-clean rates, the pipeline is altering speaker identity rather than preserving it.

Watch

Extended reading notes

Core claim

The central claim is that generative audioencoder embeddings are the right substrate for vocoder-based speech enhancement, because they carry the full speech content, including timbre and pitch, that a vocoder needs to reproduce the original speaker. With the Dasheng encoder frozen and a vocoder trained only on clean embeddings, the system reports PESQ 2.32 on the Valentini set and 2.24 on the DNS1 set, and speaker-fidelity similarity of 0.783 and 0.881, whereas WavLM and Whisper embeddings, though they suppress noise well, drop speaker similarity to about 0.41 and 0.49. In a listening study the Dasheng pipeline scores 3.87 MOS against 3.11 for Demucs, which the paper takes as evidence that denoising generative embeddings produces perceptually cleaner speech without erasing the speaker.

Load-bearing premise

The method assumes that the 768-dimensional generative audioencoder embeddings retain enough speaker and timbre information that a vocoder trained only on clean embeddings can faithfully reconstruct the original speaker from a denoised noisy embedding, and that Euclidean-distance denoising removes noise without erasing that identity.

Editorial extensions

If this is right

  • A separate, very small network can perform speech enhancement once a generative audioencoder and vocoder already exist; the ablation shows even 1.2M-parameter MLP2 produces usable perceptual quality.
  • The same pre-trained vocoder works across different denoise encoders, so scaling or swapping the denoising module does not require retraining synthesis.
  • The vocoder is trained from clean speech only, so the demand for paired noisy/clean data is confined to the lightweight denoise encoder.
  • Because generative embeddings are claimed to retain timbre, the same encoder-vocoder pair should transfer to dereverberation, source separation, and bandwidth expansion by retraining only the denoise encoder.
  • Discriminative audioencoders are a poor foundation for vocoder-based enhancement because their embeddings lose speaker identity, so future model selection should prioritize invertible or generative representations.

Reading between the lines

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

  • A useful stress test would be to pass already-clean speech through the full encode-denoise-synthesize loop; the resulting drop in MOS or speaker similarity would quantify how much distortion the vocoder itself adds.
  • The comparison across audioencoders trains a separate vocoder for each encoder, so a cleaner decomposition would train a vocoder on WavLM or Whisper embeddings with identical effort to see whether low speaker fidelity is intrinsic to those embeddings or to the synthesis pair.
  • If the embedding manifold is smooth, the ablation suggests the denoise module could shrink further; the practical floor is likely where noisy embeddings can no longer be mapped back into the vocoder's clean-embedding manifold.
  • The method's promise for other tasks depends on whether the generative embedding space is degradation-agnostic; a direct test is to train the same denoise encoder with reverberant or separated-source targets and measure whether the vocoder output follows.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes a speech enhancement system that operates in the embedding domain: a frozen pre-trained audio encoder extracts embeddings from noisy speech, a compact denoise encoder maps them toward clean-speech embeddings under an MSE loss, and a pre-trained Vocos vocoder synthesizes the waveform from the denoised embeddings. The authors compare three audio encoders (WavLM, Whisper, and Dasheng) plus a log-mel spectrogram baseline, report intrusive and non-intrusive quality metrics, speaker-fidelity results, a parameter/architecture ablation of the denoise encoder, and a subjective listening test against the Demucs enhancement model. The central claim is that the generative audio encoder Dasheng yields better perceptual quality (MOS 3.87 vs 3.11 for Demucs) and better speaker fidelity than the discriminative encoders, while the tiny denoise encoder (as small as 1.2M parameters) suffices for competitive performance.

Significance. If the claims hold, the paper contributes a simple, parameter-efficient recipe for speech enhancement that leverages frozen generative audio encoders, and it draws a useful empirical distinction between discriminative and generative embedding spaces. The study is strengthened by several concrete elements: the audio encoder and vocoder are frozen during denoise-encoder training, ablations systematically vary denoise-encoder size and architecture, a subjective listening test is included, and the authors release code, a checkpoint, and audio examples. The main value is not a new state-of-the-art result but a well-scoped demonstration that embedding-domain enhancement with a generative encoder can preserve speaker identity and perceived quality better than embedding-domain enhancement with discriminative encoders. This significance is conditional, however, on verifying that the vocoder faithfully inverts the denoised embeddings, because the current evidence leaves open the possibility that the subjective advantage comes from the vocoder's generative prior rather than from accurate reconstruction of cleaned speech.

major comments (4)
  1. [Section 2, Figure 1, Section 3.2] The load-bearing reconstruction path is not validated. Vocos is trained only on clean-speech embeddings (Figure 1c), but at inference it must synthesize from MSE-denoised embeddings of noisy speech; MSE in 768-dimensional space does not guarantee that the denoised embedding lies on the clean-embedding manifold. Please add (a) a reconstruction sanity check in which clean embeddings are fed to the trained vocoder and the output is compared to the original clean waveform with PESQ/STOI/DNSMOS, (b) a distributional analysis of denoised versus clean embeddings (e.g., nearest-neighbor distances or a simple manifold statistics report), and (c) a control ablation that feeds unprocessed noisy embeddings directly to the vocoder, so the reader can separate the contribution of the vocoder prior from the contribution of the denoise encoder. Without one of these checks, the subjective MOS advantage over Demucs could be produced by Vocos generating pleasant audio regardless of whether the denoised embedding faithfully represents the cleaned speech.
  2. [Section 4.1, Table 1] The standard intrusive metrics in Table 1 do not support the central enhancement claim. On Valentini, the Dasheng system improves PESQ over noisy (2.32 vs 1.97) but reduces STOI (0.90 vs 0.92), and both WavLM and Whisper degrade both PESQ and STOI substantially relative to noisy; on DNS1, WavLM and Whisper degrade both metrics, while Dasheng only ties STOI (0.92 vs 0.92). The text in Section 4.1 states that audioencoder-based methods show 'distinct performance improvement' in noise suppression, but that conclusion relies exclusively on non-intrusive metrics (DNSMOS, NISQAv2), and the paper does not explain why the intrusive metrics disagree so sharply. Please report per-sample statistics or significance tests for the PESQ/STOI differences, discuss the intelligibility loss as a limitation, and justify why DNSMOS/NISQAv2 should be given priority over PESQ/STOI for this system.
  3. [Section 4.3, Table 5] The subjective listening test is central to the paper's main claim but is underreported. The paper gives 17 listeners and 15 audio samples but no test methodology (e.g., MUSHRA, ITU-T P.808, or a paired-comparison protocol), no confidence intervals or statistical significance test for the 0.76 MOS gap versus Demucs, no description of sample selection or SNR distribution, and no inter-rater reliability measure. Please add these details and a significance analysis. In addition, the Demucs baseline should be justified as an appropriate 'state-of-the-art' speech enhancement system, and ideally the subjective comparison should include at least one recent SE model; the reference as written (Défossez et al., 'Real time speech enhancement in the waveform domain') is several years old.
  4. [Section 4.1, Table 4] The speaker-fidelity comparison in Table 4 is missing a critical reference: the ECAPA-TDNN cosine similarity between the clean speaker embedding and the embedding of the unprocessed noisy speech. Without this baseline, the low values for WavLM and Whisper could be caused by vocoder resynthesis artifacts or by the enhancement pipeline's signal distortion rather than by loss of speaker information in the embedding itself. Please report the speaker similarity of the noisy input and of the LMS baseline (the latter is already in the table but its relationship to the claim is not discussed), and consider also reporting a speaker-verification equal-error rate or at least error bars across the test utterances.
minor comments (6)
  1. [Section 1] There is a typo in the Introduction: 'robust auudioencoders' should be 'robust audioencoders'.
  2. [Section 3.1] The sentence 'This technique simulates noisy conditions and ensures the that the model is robust to different noise environments' contains a duplicated article; please simplify to 'ensures that the model is robust...'.
  3. [Section 3.3] The paragraph on NISQAv2 ends with the fragment 'The overall score of' before the sentence is cut off. Please complete the sentence and state explicitly whether the overall NISQAv2 score or a subscore is reported in Tables 1 and 3.
  4. [Table 1] Several cells in Table 1 have run-together numeric values (e.g., '3.493.494.133.26 4.12' in the WavLM row on Valentini). Please format all rows so that each metric column contains exactly one number.
  5. [Table 2] The color coding of fine-tuned (red) versus frozen (blue) parameters may not survive print or grayscale rendering; consider adding a symbol or footnote to distinguish these cases.
  6. [Section 4.3] The phrase 'an existing state-of-the-art SE model' in the abstract and Section 4.3 is only supported by a single, older baseline (Demucs/Denoiser); please temper the wording or add more recent baselines.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the pipeline is empirically validated against external baselines, and the self-cited Dasheng encoder is pretrained independently for audio classification.

full rationale

The paper does not present an analytic derivation whose output reduces to a fitted constant or to a self-citation chain. The method is a three-stage pipeline: a frozen pretrained audioencoder (Dasheng, WavLM, or Whisper) extracts embeddings, a compact ViT denoiser is trained with MSE between noisy and clean embeddings, and a Vocos vocoder (external, [12]) is trained on clean embeddings. Each stage is independently trained and the central claims are supported by benchmark evaluations against external systems (Demucs, WavLM, Whisper) on standard datasets (DNS1, Valentini). The Dasheng encoder is from the authors' prior work [10], but it is an open-sourced checkpoint pretrained for general audio classification, not for speech enhancement, and its suitability here is empirically tested in this paper rather than assumed. The denoiser's MSE objective is not equivalent to the evaluation metrics; in fact, Table 1 shows PESQ/STOI often below the noisy reference, demonstrating that the pipeline is not trivially forced to succeed. The skeptical concern that the Vocos vocoder may impose its own generative prior on subjective MOS scores is a validity/correctness risk, not a circularity, because no equation or fitted parameter is renamed as a prediction. The paper also openly releases inference code and audio examples, providing external falsifiability. Therefore, no load-bearing circular step is identifiable under the stated criteria.

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

This is an empirical machine learning paper with no analytic derivation. The load-bearing assumptions are about the invertibility and semantic properties of pretrained embedding spaces, not about fitted physical or mathematical constants. No new entities are postulated.

assumptions (4)
  • domain assumption Dasheng embeddings are invertible enough for a Vocos vocoder trained on clean speech embeddings to reconstruct high-quality speech.
    The entire inference pipeline depends on this; it enters in Section 2 and Figure 1c, where the vocoder is trained only on embeddings of clean speech.
  • domain assumption MSE denoising in the embedding space preserves speech content and speaker identity.
    Section 2 trains the denoise encoder with MSE between noisy and clean embeddings; the paper provides no proof that Euclidean closeness in embedding space corresponds to perceptual or speaker closeness.
  • domain assumption A vocoder trained on clean-speech embeddings transfers to denoised noisy embeddings without a domain gap.
    The vocoder never sees noisy or denoised embeddings during training, so this transfer is assumed in Sections 2 and 3.2.
  • domain assumption Subjective MOS from 17 listeners and 15 audio clips is representative enough to support the state-of-the-art comparison.
    Section 4.3 reports only mean MOS and no confidence intervals, so the subjective claim rests on the assumption that this small sample is representative.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Speech Enhancement via Embeddings from Pre-trained Generative Audioencoders." pith.science (2026). https://pith.science/paper/V5DZLA2A

@misc{pith2026250611514,
  author       = {Pith},
  title        = {Pith review of: Efficient Speech Enhancement via Embeddings from Pre-trained Generative Audioencoders},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V5DZLA2A}},
  note         = {Machine review of arXiv:2506.11514}
}
read the original abstract

Recent research has delved into speech enhancement (SE) approaches that leverage audio embeddings from pre-trained models, diverging from time-frequency masking or signal prediction techniques. This paper introduces an efficient and extensible SE method. Our approach involves initially extracting audio embeddings from noisy speech using a pre-trained audioencoder, which are then denoised by a compact encoder network. Subsequently, a vocoder synthesizes the clean speech from denoised embeddings. An ablation study substantiates the parameter efficiency of the denoise encoder with a pre-trained audioencoder and vocoder. Experimental results on both speech enhancement and speaker fidelity demonstrate that our generative audioencoder-based SE system outperforms models utilizing discriminative audioencoders. Furthermore, subjective listening tests validate that our proposed system surpasses an existing state-of-the-art SE model in terms of perceptual quality.

Figures

Figures reproduced from arXiv: 2506.11514 by the authors.

Figure 1
Figure 1. The proposed system: a. illustrates the overall framework, which enhances speech by performing denoising at the embedding level. b. depicts the training process of the proposed denoising encoder, implemented as a linear layer. c. presents the training process of the vocoder. ditionally, subjective listening tests indicate that our proposed system outperforms an existing state-of-the-art SE model. 2. Proposed Approac… view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. UniPASE: A Generative Model for Universal Speech Enhancement with High Fidelity and Low Hallucinations

    eess.AS 2026-04 unverdicted novelty 6.0 of 10

    UniPASE extends the PASE framework with DeWavLM-Omni to convert degraded speech into high-fidelity, low-hallucination audio across sampling rates via phonetic enhancement, acoustic adaptation, and multi-rate vocoding.

Reference graph

Works this paper leans on

32 extracted references · 26 canonical work pages · cited by 1 Pith paper

  1. [6]

    Sixty years of frequency-domain monaural speech en- hancement: From traditional to deep learning methods,

    C. Zheng, H. Zhang, W. Liu, X. Luo, A. Li, X. Li, and B. C. J. Moore, “Sixty years of frequency-domain monaural speech en- hancement: From traditional to deep learning methods,”Trends in Hearing, vol. 27, 2023

  2. [8]

    Investigating self-supervised learning for speech enhancement and separation,

    Z. Huang, S. Watanabe, S.-W. Yang, L. P. Garc ´ıa-Perera, and S. Khudanpur, “Investigating self-supervised learning for speech enhancement and separation,”ICASSP 2022 - 2022 IEEE Inter- national Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 6837–6841, 2022

  3. [1]

    Most deep neu- ral network (DNN) based SE methods primarily rely on mask or signal prediction techniques with a deep-learning framework tailored towards the SE task [1]

    Introduction Speech enhancement (SE) is a fundamental task in audio sig- nal processing, aiming to improve the quality and intelligibility of speech signals that are degraded by noise. Most deep neu- ral network (DNN) based SE methods primarily rely on mask or signal prediction techniques with a deep-learning framework tailored towards the SE task [1]. Re...

  4. [2]

    Initially, the noisy speech is in- put into a pre-trained audioencoder, yielding a noisy embed- ding

    Proposed Approach As depicted in Figure 1a, our SE inference process is structured into three distinct modules. Initially, the noisy speech is in- put into a pre-trained audioencoder, yielding a noisy embed- ding. Subsequently, the denoised embedding is extracted from the noisy embedding via a denoise encoder. Ultimately, the de- noised embedding is emplo...

  5. [3]

    Experiments 3.1. Data The training of our system leverages the clean speech and noise dataset from the ICASSP 2022 Deep Noise Suppression (DNS4) Challenge [16], with additional clean speech data from the Common V oice project [17] to augment the training cor- pus. In the vocoder training process, we exclusively employ clean speech to train the model in a ...

  6. [4]

    Evaluation of Different Audioencoders Table 1 presents the evaluation results for both Valentini and the DNS1 test sets

    Results 4.1. Evaluation of Different Audioencoders Table 1 presents the evaluation results for both Valentini and the DNS1 test sets. When considering intrusive metrics such as PESQ and STOI, the performance of all methods is decreased over the noisy signal, with the exception of the PESQ scores for Dasheng and LMS in the DNS1 test set. In contrast, non- ...

  7. [5]

    However, we chose not to im- plement global fine-tuning due to several compelling considera- tions

    Discussion and Conclusion Global fine-tuning of all parameters within our SE system might yield performance improvement. However, we chose not to im- plement global fine-tuning due to several compelling considera- tions. Primarily, by utilizing a pre-trained audioencoder, we can leverage pre-computed embeddings from other tasks without in- curring additio...

  8. [7]

    Wavlm: Large-scale self- supervised pre-training for full stack speech processing,

    S. Chen, C. Wang, Z. Chen, Y . Wu, S. Liu, Z. Chen, J. Li, N. Kanda, T. Yoshioka, X. Xiao, J. Wu, L. Zhou, S. Ren, Y . Qian, Y . Qian, M. Zeng, and F. Wei, “Wavlm: Large-scale self- supervised pre-training for full stack speech processing,”IEEE Journal of Selected Topics in Signal Processing, vol. 16, pp. 1505–1518, 2021

Show all 32 references
  1. [9]

    Boosting self-supervised embeddings for speech en- hancement,

    K.-H. Hung, S.-W. Fu, H.-H. Tseng, H.-T. Chiang, Y . Tsao, and C.-W. Lin, “Boosting self-supervised embeddings for speech en- hancement,” inInterspeech, 2022

  2. [10]

    Hifi-gan-2: Studio-quality speech enhancement via generative adversarial networks condi- tioned on acoustic features,

    J. Su, Z. Jin, and A. Finkelstein, “Hifi-gan-2: Studio-quality speech enhancement via generative adversarial networks condi- tioned on acoustic features,”2021 IEEE Workshop on Applica- tions of Signal Processing to Audio and Acoustics (WASPAA), pp. 166–170, 2021

  3. [11]

    Self- supervised learning for speech enhancement through synthesis,

    B. Irvin, M. Stamenovic, M. Kegler, and L.-C. Yang, “Self- supervised learning for speech enhancement through synthesis,” ICASSP 2023 - 2023 IEEE International Conference on Acous- tics, Speech and Signal Processing (ICASSP), pp. 1–5, 2022

  4. [12]

    Speech enhancement using self-supervised pre-trained model and vector quantization,

    X.-Y . Zhao, Q. shi Zhu, and J. Zhang, “Speech enhancement using self-supervised pre-trained model and vector quantization,”2022 Asia-Pacific Signal and Information Processing Association An- nual Summit and Conference (APSIPA ASC), pp. 330–334, 2022

  5. [13]

    Spectrum-aware neural vocoder based on self-supervised learning for speech en- hancement,

    Y . Song, D. Kim, H.-G. Kang, and N. Madhu, “Spectrum-aware neural vocoder based on self-supervised learning for speech en- hancement,”2024 32nd European Signal Processing Conference (EUSIPCO), pp. 16–20, 2024

  6. [14]

    Robust speech recognition via large-scale weak su- pervision,

    A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust speech recognition via large-scale weak su- pervision,” inInternational Conference on Machine Learning, 2022

  7. [15]

    Scaling up masked audio encoder learning for general audio clas- sification,

    H. Dinkel, Z. Yan, Y . Wang, J. Zhang, Y . Wang, and B. Wang, “Scaling up masked audio encoder learning for general audio clas- sification,”Interspeech, p. 547–551, 2024

  8. [16]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,”ICLR, 2021

  9. [17]

    V ocos: Closing the gap between time-domain and fourier-based neural vocoders for high-quality audio synthesis,

    H. Siuzdak, “V ocos: Closing the gap between time-domain and fourier-based neural vocoders for high-quality audio synthesis,” arXiv preprint arXiv:2306.00814, 2023

  10. [18]

    A convnet for the 2020s,

    Z. Liu, H. Mao, C. Wu, C. Feichtenhofer, T. Darrell, and S. Xie, “A convnet for the 2020s,”2022 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pp. 11 966–11 976, 2022

  11. [19]

    Hifi-gan: Generative adversarial networks for efficient and high fidelity speech synthesis,

    J. Kong, J. Kim, and J. Bae, “Hifi-gan: Generative adversarial networks for efficient and high fidelity speech synthesis,”Ad- vances in Neural Information Processing Systems, vol. 33, p. 17022–17033, 2020

  12. [20]

    Univnet: A neural vocoder with multi-resolution spectrogram discriminators for high-fidelity waveform generation,

    W. Jang, D. C. Y . Lim, J. Yoon, B. Kim, and J. Kim, “Univnet: A neural vocoder with multi-resolution spectrogram discriminators for high-fidelity waveform generation,” inInterspeech, 2021

  13. [21]

    Icassp 2022 deep noise suppression challenge,

    H. Dubey, V . Gopal, R. Cutler, S. Matusevych, S. Braun, E. S. Eskimez, M. Thakker, T. Yoshioka, H. Gamper, and R. Aichner, “Icassp 2022 deep noise suppression challenge,” inICASSP, 2022

  14. [22]

    Common voice: A massively-multilingual speech corpus,

    R. Ardila, M. Branson, K. Davis, M. Henretty, M. Kohler, J. Meyer, R. Morais, L. Saunders, F. M. Tyers, and G. Weber, “Common voice: A massively-multilingual speech corpus,” in Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), 2020, pp. 4211–4215

  15. [23]

    The interspeech 2020 deep noise suppression challenge: Datasets, subjective testing framework, and challenge results,

    C. K. Reddy, V . Gopal, R. Cutler, E. Beyrami, R. Cheng, H. Dubey, S. Matusevych, R. Aichner, A. Aazami, S. Braun et al., “The interspeech 2020 deep noise suppression challenge: Datasets, subjective testing framework, and challenge results,” in INTERSPEECH, 2020

  16. [24]

    Noisy speech database for training speech enhancement algorithms and tts models,

    C. Valentini-Botinhao, “Noisy speech database for training speech enhancement algorithms and tts models,” 2017

  17. [25]

    An al- gorithm for intelligibility prediction of time–frequency weighted noisy speech,

    C. H. Taal, R. C. Hendriks, R. Heusdens, and J. R. Jensen, “An al- gorithm for intelligibility prediction of time–frequency weighted noisy speech,”IEEE Transactions on Audio, Speech, and Lan- guage Processing, vol. 19, pp. 2125–2136, 2011

  18. [26]

    Perceptual eval- uation of speech quality (pesq)-a new method for speech quality assessment of telephone networks and codecs,

    A. Rix, J. Beerends, M. Hollier, and A. Hekstra, “Perceptual eval- uation of speech quality (pesq)-a new method for speech quality assessment of telephone networks and codecs,” in2001 IEEE In- ternational Conference on Acoustics, Speech, and Signal Process- ing, vol. 2, 2001, ...

  19. [27]

    Dnsmos p.835: A non- intrusive perceptual objective speech quality metric to evaluate noise suppressors,

    C. K. Reddy, V . Gopal, and R. Cutler, “Dnsmos p.835: A non- intrusive perceptual objective speech quality metric to evaluate noise suppressors,” inICASSP, 2022

  20. [28]

    Nisqa: A deep cnn-self-attention model for multidimensional speech quality pre- diction with crowdsourced datasets,

    G. Mittag, B. Naderi, A. Chehadi, and S. M ¨oller, “Nisqa: A deep cnn-self-attention model for multidimensional speech quality pre- diction with crowdsourced datasets,” inInterspeech, 2021

  21. [29]

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

    B. Desplanques, J. Thienpondt, and K. Demuynck, “Ecapa-tdnn: Emphasized channel attention, propagation and aggregation in tdnn based speaker verification,” inInterspeech, 2020

  22. [30]

    Speechbrain: A general-purpose speech toolkit,

    M. Ravanelli, T. Parcollet, and P. W. V . P. et. al., “Speechbrain: A general-purpose speech toolkit,”ArXiv, vol. abs/2106.04624, 2021

  23. [31]

    Real time speech en- hancement in the waveform domain,

    A. D ´efossez, G. Synnaeve, and Y . Adi, “Real time speech en- hancement in the waveform domain,” 2020

  24. [32]

    P.808 : Subjective evaluation of speech quality with a crowdsourcing approach,

    ITU-R, “P.808 : Subjective evaluation of speech quality with a crowdsourcing approach,” 2015

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.