Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Neural Vocoders as Speech Enhancers

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Speech enhancement and neural vocoding are opposite rank-restoration directions of the same task, and one jointly trained network can handle both.

desk verdict Solid empirical result that SE networks can vocode and a single model can do both, with a rank-restoration story that is correlational, not causal; worth refereeing after the mechanism claim is toned down and statistics are added. read the letter →

arxiv 2501.13465 v1 pith:V3NP4IYA submitted 2025-01-23 cs.SD eess.AS

Andong Li , Zhihang Sun , Fengyuan Hao , Xiaodong Li , Chengshi Zheng This is my paper · ORCID
classification cs.SDeess.AS
keywords neuralvocoderspeechenhancementspectralrankjointtrainingrestorationmel-spectrogramband-splitRNNpseudo-inversemelfilter
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

This paper argues that speech enhancement and neural vocoding are opposite directions of a single speech-restoration process, not unrelated tasks. The authors measure the rank of magnitude spectrograms and find that additive noise pushes rank upward while the mel transform followed by its pseudo-inverse pushes rank downward. They then show that an enhancement network trained to lower rank can be retrained to raise rank: on the LJSpeech benchmark, the BSRNN-L enhancement model reports the best vocoding scores among the compared vocoders, and a jointly trained BSRNN-M handles both denoising and vocoding with performance comparable to single-task models. The paper proposes that both tasks be viewed as restoring the low-rank spectral structure of clean speech.

What carries the argument

The mechanism is the spectral rank restoration trajectory: the rank of the magnitude spectrogram matrix $\mathbf{R}(\cdot)$ goes down under the Linear2Mel transform and its pseudo-inverse, and up under additive noise. The rank bounds are $\mathbf{R}(|X|) \leq \mathbf{R}(|S|)+\mathbf{R}(|N|)$ and $\mathbf{R}(\hat{Y}) \leq \min\{\mathbf{R}(|S|),\mathbf{R}(AA^+)\}$, with $A$ the Linear2Mel matrix and $A^+$ its pseudo-inverse. The pseudo-inverse is the key input adaptation: it projects mel features back to a linear-scale spectrogram of the same shape as a noisy input, so the same SE network can ingest either degradation. The joint-training procedure randomly chooses denoising or vocoding at each step with probability $p=0.5$, and the ablation studies select log-masking plus Griffin-Lim proxy phase as the settings that make transfer work best.

What would settle it

Train the same BSRNN architecture on two synthetic corruptions of equal severity that differ only in rank sign — one additive-noise corruption that raises rank and one mel-compression corruption that lowers it — then test cross-task transfer; if transfer is symmetric or absent in the predicted direction, the rank-restoration trajectory is not the operative mechanism.

Watch

Extended reading notes

Core claim

The central claim is that existing speech enhancement models can be converted into neural vocoders, and that a single jointly trained model can do both tasks. The evidence is threefold. First, rank analysis on the VoiceBank-Demand corpus shows that noisy inputs have positive rank difference relative to clean speech while mel-compressed inputs reconstructed through the pseudo-inverse have negative rank difference. Second, using log-masking for magnitude estimation and the pseudo-inverse mel-to-linear projection as input, BSRNN variants vocode at or above the level of dedicated vocoders, with BSRNN-L reaching the best scores on LJSpeech among all compared systems. Third, joint training that randomly samples denoising and vocoding with equal probability yields a model whose enhancement performance matches single-task training and whose vocoding performance reaches single-task level by 1.75 million steps. The paper concludes that the two tasks differ only in the spectral rank restoration trajectory, making room for a unified speech-restoration framework.

Load-bearing premise

The whole unification rests on the premise that the only meaningful difference between denoising and vocoding is the direction of spectral-rank change; if what really transfers between tasks is general network capacity or shared low-level speech structure, the explanation collapses even if the benchmark results stand.

Editorial extensions

If this is right

  • Existing time-frequency domain speech enhancement networks can be trained as neural vocoders with competitive or better quality than dedicated vocoders; BSRNN-L reports the best scores on LJSpeech among the compared systems.
  • A single BSRNN-M jointly trained on denoising and vocoding with balanced task sampling matches single-task models: enhancement performance is on par, and vocoding quality catches up by 1.75 million training steps.
  • Log-masking for magnitude estimation and Griffin-Lim proxy phase for time-domain SE inputs are the configurations that make the transfer work best.
  • The rank framework extends the idea of speech restoration: other front-end tasks with opposite rank trajectories, such as dereverberation and bandwidth extension, could be folded into the same joint-training scheme.

Reading between the lines

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

  • The paper's title points to the reverse direction, but the experiments run enhancement models as vocoders, not vocoder models as enhancers; a symmetric test of a vocoder trained on noisy input to denoise would test whether the unification is bidirectional.
  • The rank story is correlational: no experiment manipulates spectral rank directly. An intervention that projects internal representations to a target rank, or that uses corruptions matched in distortion but opposite in rank sign, would turn the explanation into a testable mechanism.
  • If the unification holds, one deployed model could replace two model families — text-to-speech backends and denoising front-ends — and other front-end degradations become natural candidates for the same joint-training recipe.
  • The 1.75-million-step convergence suggests the joint model is not yet saturated; scaling the training mixture with more vocoded and degraded corpora, as the paper anticipates, may push both tasks further.
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

3 major / 6 minor

Summary. This paper proposes that speech enhancement (denoising) and neural vocoding can be unified through the lens of spectral rank manipulation: it argues that additive noise tends to increase spectral rank while mel-compression followed by pseudo-inverse tends to decrease it. Based on this rank analysis, the authors hypothesize that existing SE models should be adaptable to vocoding and that a single model can handle both tasks. They empirically show that SE networks (BSRNN, ConvTasNet, GCRN, HD-Demucs) can be trained as vocoders, with BSRNN-L achieving the best scores on LJSpeech, and that a jointly trained BSRNN-M reaches performance comparable to single-task models. Code is made available.

Significance. If the empirical claims hold, the paper makes a useful contribution by demonstrating that SE architectures can be repurposed as vocoders and that joint training yields a single model for both tasks. The authors make a strong effort toward fairness by retraining all baselines on a common data split. The rank perspective is conceptually appealing but, as presented, is a heuristic: the paper does not establish a causal link between rank behavior and the success of transfer. Nevertheless, the empirical core is valuable and likely reproducible, and the released code is a strength.

major comments (3)
  1. [Sec. II, Eqs. (4)-(5), Fig. 2] Eq. (4) provides only an upper bound R(|S|+|N|) ≤ R(|S|)+R(|N|), which does not establish that adding noise tends to increase rank. Moreover, exact matrix rank of real-valued spectrograms is almost surely full, so the histograms in Fig. 2 must rely on a thresholded rank with η=0.5; the connection between this thresholded rank and Eq. (4) is not shown. The claim that noise and mel-compression move rank in opposite directions is therefore not proven by the stated mathematics, and the histograms alone do not close that gap.
  2. [Sec. III] The sentence 'the difference between the two tasks lie only in the spectral rank restoration trajectory' makes a strong causal claim. The paper never intervenes on rank (e.g., via rank-constrained training or artificially rank-modified inputs), nor does it measure whether model outputs actually move toward the target rank. Alternative explanations for the successful transfer—such as shared spectral structure or the pseudo-inverse mel input making vocoding a spectral-mapping problem (FreeV, [15])—are not ruled out. The empirical unification may survive, but the advertised 'rank manipulation perspective' is not supported.
  3. [Table III and Fig. 3] The 'new state-of-the-art' claim rests on one benchmark (LJSpeech), one run per model, and no error bars or significance tests. Fig. 3 reports single runs and uses phrases such as 'no significant difference' and 'surpassed' without statistical evidence. Given the small metric gaps between some models (e.g., BSRNN-M and FreeV in Table III), reporting variance over multiple runs or paired significance tests is necessary to support the strength of the claims.
minor comments (6)
  1. [Sec. II, Eq. (7)] In the discussion after Eq. (7), the text says 'mel-oriented degradation is associated with a negative rank difference, i.e., ∆Rse ≤ 0'; this appears to be a typo and should refer to ∆Rvo.
  2. [Sec. IV-A] The LJSpeech split is described as following the VITS repository's guidelines, but no URL or explicit split file is provided; please include a reference or link for reproducibility.
  3. [Fig. 2] The caption states that ranks are calculated from the Voicebank-Demand training set, but the experiments use LJSpeech and LibriTTS; clarify whether the rank analysis is meant as a general observation or specific to the evaluation datasets.
  4. [Algorithm 1] The task sampling probability p=0.5 is a free parameter; the paper does not study sensitivity to p, which could affect the joint-training conclusions.
  5. [Table III] The table reports V/UV F1, Periodicity, Pitch-RMSE, and F0-RMSE; definitions of these metrics and their computation details should be provided or cited in the experimental section.
  6. [General] There are several minor typos, including 'V oicebank' in Fig. 2, 'V ocos' in Table III, 'reposity' in Sec. IV-A, and 'speech vocoding task. employing' in Sec. V-B.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the empirical comparisons are external and the rank hypothesis, though under-supported, does not enter the training or evaluation as a fitted input.

full rationale

The central empirical claims are tested against external baselines (HiFiGAN, iSTFTNet, APNet, APNet2, Vocos, FreeV, and four SE models) on LJSpeech and LibriTTS+Noise, with all vocoder baselines retrained under the same split. No parameter is fitted to a subset of the evaluation data and then renamed as a prediction; the only 'prediction' framing is the spectral-rank hypothesis, which is motivational and never enters the loss, training objective, or metrics. Equations (4) and (5) are inequalities used to motivate rank trends; they are not used to construct the unified model, and the experimental success does not reduce to them. The assertion in Sec. III that the two tasks 'lie only in the spectral rank restoration trajectory' is an under-supported overclaim (the upper bound does not prove rank increase, and the Fig. 2 histograms use a threshold-dependent rank), but that is a correctness or evidence-strength concern, not circularity. Self-citations [32]--[34] appear only in a side remark about masking versus mapping and are not load-bearing for the central claim. The joint-training comparison fairly reports that the single model needs about 1.75M total steps (roughly 875k per task) to match a 1M-step single-task vocoder, so no forced equivalence is hidden.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

No new physical entities. The rank manipulation framework is a conceptual relabeling, not a new mechanism. The free parameters listed are choices in the analysis and training procedure, not fitted constants.

free parameters (2)
  • Rank threshold eta = 0.5
    Used to compute spectral rank in Fig. 2; histograms and the claimed sign of rank differences depend on this choice, with no sensitivity analysis.
  • Task sampling probability p = 0.5
    In joint training, p=0.5 alternates denoising and vocoding per step, which halves effective per-task steps; the comparison at fixed global steps is therefore partly confounded.
assumptions (6)
  • standard math Matrix rank subadditivity and rank of a product (Eqs. 4-5)
    Used to argue noise tends to increase rank and mel compression tends to decrease it.
  • domain assumption Pseudo-inverse of the Linear2Mel transform gives a valid linear-scale spectrogram approximation (FreeV [15])
    Vocoding input is formed as Y A+; the quality of this inverse is inherited from prior work.
  • domain assumption Clean speech spectrograms have a naturally low rank related to speech quality [16]
    Underpins the goal of restoring the natural characteristic spectral rank.
  • domain assumption Phase can be omitted in rank analysis because rank is associated with eigenvalues tied to spectral magnitude
    Eqs. (4)-(5) drop phase; the justification is qualitative.
  • domain assumption Environmental noise is uncorrelated with speech and therefore tends to raise rank rather than cancel spectral components
    Converts an upper bound in Eq. (4) into the empirical claim DeltaRse >= 0.
  • ad hoc to paper A single architecture with one parameter set can learn both rank-increasing and rank-decreasing mappings
    This is the tested hypothesis itself, introduced in Sec. III as the basis for joint training.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neural Vocoders as Speech Enhancers." pith.science (2026). https://pith.science/paper/V3NP4IYA

@misc{pith2026250113465,
  author       = {Pith},
  title        = {Pith review of: Neural Vocoders as Speech Enhancers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V3NP4IYA}},
  note         = {Machine review of arXiv:2501.13465}
}
read the original abstract

Speech enhancement (SE) and neural vocoding are traditionally viewed as separate tasks. In this work, we observe them under a common thread: the rank behavior of these processes. This observation prompts two key questions: \textit{Can a model designed for one task's rank degradation be adapted for the other?} and \textit{Is it possible to address both tasks using a unified model?} Our empirical findings demonstrate that existing speech enhancement models can be successfully trained to perform vocoding tasks, and a single model, when jointly trained, can effectively handle both tasks with performance comparable to separately trained models. These results suggest that speech enhancement and neural vocoding can be unified under a broader framework of speech restoration. Code: https://github.com/Andong-Li-speech/Neural-Vocoders-as-Speech-Enhancers.

Figures

Figures reproduced from arXiv: 2501.13465 by the authors.

Figure 1
Figure 1. Illustrations of the signal degradation process [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Relative rank difference w.r.t. target spectrum of denoising and vocoding tasks. The ranks are calculated from the training set of Voicebank￾Demand benchmark [17]. The absolute threshold η is set to 0.5 for rank calculation and better visualization. manipulation, potentially enabling them to perform either SE/vocoding task or both. We empirically demonstrate that 1) existing SE models can successfully perform vocodi… view at source ↗
Figure 3
Figure 3. Metric comparisons for joint denoising-vocoding task and its single [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] 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. Learning Neural Vocoder from Range-Null Space Decomposition

    cs.SD 2025-07 conditional novelty 5.0 of 10

    By decomposing spectrogram reconstruction into a fixed pseudo-inverse range-space step and a learned null-space detail step, RNDVoC reaches near-BigVGAN quality with about 3% of the parameters and a 10x CPU speed-up.

Reference graph

Works this paper leans on

40 extracted references · 31 canonical work pages · cited by 1 Pith paper

  1. [15]

    FreeV: Free Lunch For Vocoders Through Pseudo Inversed Mel Filter,

    Y . Lv, H. Li, Y . Yan, J. Liu, D. Xie, and L. Xie, “FreeV: Free Lunch For Vocoders Through Pseudo Inversed Mel Filter,” arXiv preprint arXiv:2406.08196, 2024

    arXiv 2024

  2. [1]

    Tacotron: Towards End- to-End Speech Synthesis,

    Y . Wang, R. Skerry-Ryan, D. Stanton, Y . Wu, R. J. Weiss, N. Jaitly, Z. Yang, Y . Xiao, Z. Chen, S. Bengio et al., “Tacotron: Towards End- to-End Speech Synthesis,” in Proc. Interspeech. ISCA, 2017, pp. 4006– 4010

    work page 2017

  3. [2]

    Naturalspeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers,

    K. Shen, Z. Ju, X. Tan, E. Liu, Y . Leng, L. He, T. Qin, J. Bian et al., “Naturalspeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers,” in Proc. ICLR

    work page

  4. [3]

    Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions,

    J. Shen, R. Pang, R. J. Weiss, M. Schuster, N. Jaitly, Z. Yang, Z. Chen, Y . Zhang, Y . Wang, R. Skerrv-Ryan et al. , “Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions,” in Proc. ICASSP. IEEE, 2018, pp. 4779–4783

    work page 2018

  5. [4]

    Text-to-audio generation using instruction-tuned llm and latent diffusion model,

    D. Ghosal, N. Majumder, A. Mehrish, and S. Poria, “Text-to-audio generation using instruction-tuned llm and latent diffusion model,” arXiv preprint arXiv:2304.13731, 2023

    arXiv 2023

  6. [5]

    Audit: Audio editing by following instructions with latent diffusion models,

    Y . Wang, Z. Ju, X. Tan, L. He, Z. Wu, J. Bian et al. , “Audit: Audio editing by following instructions with latent diffusion models,” in Proc. NeurIPS, vol. 36, 2023, pp. 71 340–71 357

    work page 2023

  7. [6]

    V oicefixer: Toward general speech restoration with neural vocoder,

    H. Liu, Q. Kong, Q. Tian, Y . Zhao, D. Wang, C. Huang, and Y . Wang, “V oicefixer: Toward general speech restoration with neural vocoder,” in Proc. Interspeech, 2022, p. 4232–4236

    work page 2022

  8. [7]

    Mel-FullSubNet: Mel-Spectrogram Enhancement for Improving Both Speech Quality and Asr,

    R. Zhou, X. Li, Y . Fang, and X. Li, “Mel-FullSubNet: Mel-Spectrogram Enhancement for Improving Both Speech Quality and Asr,” arXiv preprint arXiv:2402.13511, 2024

    arXiv 2024

Show all 40 references
  1. [8]

    Wavenet: A gener- ative model for raw audio,

    A. v. d. Oord, S. Dieleman, H. Zen, K. Simonyan, O. Vinyals, A. Graves, N. Kalchbrenner, A. Senior, and K. Kavukcuoglu, “Wavenet: A gener- ative model for raw audio,” arXiv preprint arXiv:1609.03499 , 2016

    2016 arXiv

  2. [9]

    SampleRNN: An Unconditional End-to- End Neural Audio Generation Model,

    S. Mehri, K. Kumar, I. Gulrajani, R. Kumar, S. Jain, J. Sotelo, A. Courville, and Y . Bengio, “SampleRNN: An Unconditional End-to- End Neural Audio Generation Model,” in Proc. ICLR, 2022

    2022

  3. [10]

    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,” in Proc. NeurIPS , vol. 33, 2020, pp. 17 022–17 033

    2020

  4. [11]

    iSTFTNet: Fast and lightweight mel-spectrogram vocoder incorporating inverse short-time Fourier transform,

    T. Kaneko, K. Tanaka, H. Kameoka, and S. Seki, “iSTFTNet: Fast and lightweight mel-spectrogram vocoder incorporating inverse short-time Fourier transform,” in Proc. ICASSP. IEEE, 2022, pp. 6207–6211

    2022

  5. [12]

    APNet: An all-frame-level neural vocoder incor- porating direct prediction of amplitude and phase spectra,

    Y . Ai and Z.-H. Ling, “APNet: An all-frame-level neural vocoder incor- porating direct prediction of amplitude and phase spectra,” IEEE/ACM Trans. Audio, Speech, Lang. Process. , vol. 31, pp. 2145–2157, 2023

    2023

  6. [13]

    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,” in Proc. ICLR

  7. [14]

    APNet2: High-Quality and High-Efficiency Neural Vocoder with Direct Prediction of Amplitude and Phase Spectra,

    H.-P. Du, Y .-X. Lu, Y . Ai, and Z.-H. Ling, “APNet2: High-Quality and High-Efficiency Neural Vocoder with Direct Prediction of Amplitude and Phase Spectra,” in Proc. NCMMSC. Springer, 2023, pp. 66–80

    2023

  8. [16]

    Speech dereverberation using non- negative convolutive transfer function and spectro-temporal modeling,

    N. Mohammadiha and S. Doclo, “Speech dereverberation using non- negative convolutive transfer function and spectro-temporal modeling,” IEEE/ACM Trans. Audio, Speech, Lang. Process. , vol. 24, no. 2, pp. 276–289, 2015

    2015

  9. [17]

    The voice bank corpus: Design, collection and data analysis of a large regional accent speech database,

    C. Veaux, J. Yamagishi, and S. King, “The voice bank corpus: Design, collection and data analysis of a large regional accent speech database,” in Proc. O-COCOSDA/CASLRE. IEEE, 2013, pp. 1–4

    2013

  10. [18]

    On training targets for su- pervised speech separation,

    Y . Wang, A. Narayanan, and D. Wang, “On training targets for su- pervised speech separation,” IEEE/ACM Trans. Audio, Speech, Lang. Process., vol. 22, no. 12, pp. 1849–1858, 2014

    2014

  11. [19]

    Signal estimation from modified short-time Fourier transform,

    D. Griffin and J. Lim, “Signal estimation from modified short-time Fourier transform,” IEEE Trans. Audio, Speech, Signal Process., vol. 32, no. 2, pp. 236–243, 1984

    1984

  12. [20]

    Libritts: A Corpus Derived from LibriSpeech for Text-to- Speech,

    H. Zen, V . Dang, R. Clark, Y . Zhang, R. J. Weiss, Y . Jia, Z. Chen, and Y . Wu, “Libritts: A Corpus Derived from LibriSpeech for Text-to- Speech,” in Proc. Interspeech, 2019, pp. 1526–1530

    2019

  13. [21]

    ICASSP 2023 deep noise suppression challenge,

    H. Dubey, A. Aazami, V . Gopal, B. Naderi, S. Braun, R. Cutler, A. Ju, M. Zohourian, M. Tang, M. Golestaneh et al. , “ICASSP 2023 deep noise suppression challenge,” IEEE Open J. Signal Process. , vol. 5, pp. 1849–1858, 2024

    2023

  14. [22]

    MUSAN: A music, speech, and noise corpus,

    D. Snyder, G. Chen, and D. Povey, “MUSAN: A music, speech, and noise corpus,” arXiv preprint arXiv:1510.08484 , 2015

    2015 arXiv

  15. [23]

    Hd-demucs: General speech restoration with heterogeneous decoders,

    D. Kim, S.-W. Chung, H. Han, Y . Ji, and H.-G. Kang, “Hd-demucs: General speech restoration with heterogeneous decoders,” in Proc. Interspeech, 2023, pp. 3829–3833

    2023

  16. [24]

    Conv-TasNet: Surpassing ideal time– frequency magnitude masking for speech separation,

    Y . Luo and N. Mesgarani, “Conv-TasNet: Surpassing ideal time– frequency magnitude masking for speech separation,” IEEE/ACM Trans. Audio, Speech, Lang. Process. , vol. 27, no. 8, pp. 1256–1266, 2019

    2019

  17. [25]

    Learning complex spectral mapping with gated convolutional recurrent networks for monaural speech enhancement,

    K. Tan and D. Wang, “Learning complex spectral mapping with gated convolutional recurrent networks for monaural speech enhancement,” IEEE/ACM Trans. Audio, Speech, Lang. Process., vol. 28, pp. 380–390, 2019

    2019

  18. [26]

    Music source separation with band-split RNN,

    Y . Luo and J. Yu, “Music source separation with band-split RNN,” IEEE/ACM Trans. Audio, Speech, Lang. Process. , vol. 31, pp. 1893– 1901, 2023

    1901

  19. [27]

    Bandwidth extension for hierarchical speech and audio coding in itu-t rec. g. 729.1,

    B. Geiser, P. Jax, P. Vary, H. Taddei, S. Schandl, M. Gartner, C. Guil- laum´e, and S. Ragot, “Bandwidth extension for hierarchical speech and audio coding in itu-t rec. g. 729.1,” IEEE Trans. Audio, Speech, Lang. Process., vol. 15, no. 8, pp. 2496–2509, 2007

    2007

  20. [28]

    A short- time objective intelligibility measure for time-frequency weighted noisy speech,

    C. H. Taal, R. C. Hendriks, R. Heusdens, and J. Jensen, “A short- time objective intelligibility measure for time-frequency weighted noisy speech,” in Proc. ICASSP. IEEE, 2010, pp. 4214–4217

    2010

  21. [29]

    UTMOS: Utokyo-Sarulab System for V oiceMOS Chal- lenge 2022,

    T. Saeki, D. Xin, W. Nakata, T. Koriyama, S. Takamichi, and H. Saruwatari, “UTMOS: Utokyo-Sarulab System for V oiceMOS Chal- lenge 2022,” Proc. Interspeech, pp. 4521–4525, 2022

    2022

  22. [30]

    PHASEN: A phase-and- harmonics-aware speech enhancement network,

    D. Yin, C. Luo, Z. Xiong, and W. Zeng, “PHASEN: A phase-and- harmonics-aware speech enhancement network,” in Proc. AAAI, vol. 34, no. 05, 2020, pp. 9458–9465

    2020

  23. [31]

    TF-GridNet: Making time-frequency domain models great again for monaural speaker separation,

    Z.-Q. Wang, S. Cornell, S. Choi, Y . Lee, B.-Y . Kim, and S. Watanabe, “TF-GridNet: Making time-frequency domain models great again for monaural speaker separation,” in Proc. ICASSP. IEEE, 2023, pp. 1–5

    2023

  24. [32]

    Taylor, can you hear me now? a taylor-unfolding framework for monaural speech enhancement,

    A. Li, S. You, G. Yu, C. Zheng, and X. Li, “Taylor, can you hear me now? a taylor-unfolding framework for monaural speech enhancement,” in Proc. IJCAI, 2022, pp. 4193–4200

    2022

  25. [33]

    Two heads are better than one: A two-stage complex spectral mapping approach for monaural speech enhancement,

    A. Li, W. Liu, C. Zheng, C. Fan, and X. Li, “Two heads are better than one: A two-stage complex spectral mapping approach for monaural speech enhancement,” IEEE/ACM Trans. Audio, Speech, Lang. Process., vol. 29, pp. 1829–1843, 2021

    2021

  26. [34]

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

    C. Zheng, H. Zhang, W. Liu, X. Luo, A. Li, X. Li, and B. C. Moore, “Sixty years of frequency-domain monaural speech enhance- ment: From traditional to deep learning methods,” Trends Hear., vol. 27, p. 23312165231209913, 2023

    2023

  27. [35]

    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,” in Proc. ICASSP. IEEE, 2022, pp. 886–890

    2022

  28. [36]

    ViSQOL v3: An open source production ready objective speech and audio metric,

    M. Chinen, F. S. Lim, J. Skoglund, N. Gureev, F. O’Gorman, and A. Hines, “ViSQOL v3: An open source production ready objective speech and audio metric,” in Proc. QoMEX. IEEE, 2020, pp. 1–6

    2020

  29. [37]

    Wavefake: A data set to facilitate audio deepfake detection,

    J. Frank and L. Sch ¨onherr, “Wavefake: A data set to facilitate audio deepfake detection,” arXiv preprint arXiv:2111.02813 , 2021

    2021 arXiv

  30. [38]

    ASVspoof 2019: Future horizons in spoofed and fake audio detection,

    M. Todisco, X. Wang, V . Vestman, M. Sahidullah, H. Delgado, A. Nautsch, J. Yamagishi, N. Evans, T. Kinnunen, and K. A. Lee, “ASVspoof 2019: Future horizons in spoofed and fake audio detection,” arXiv preprint arXiv:1904.05441 , 2019

    2019 arXiv

  31. [39]

    Singfake: Singing voice deepfake detection,

    Y . Zang, Y . Zhang, M. Heydari, and Z. Duan, “Singfake: Singing voice deepfake detection,” in Proc. ICASSP. IEEE, 2024, pp. 12 156–12 160

    2024

  32. [40]

    CtrSVDD: A Benchmark Dataset and Baseline Analysis for Controlled Singing Voice Deepfake Detection,

    Y . Zang, J. Shi, Y . Zhang, R. Yamamoto, J. Han, Y . Tang, S. Xu, W. Zhao, J. Guo, T. Toda et al., “CtrSVDD: A Benchmark Dataset and Baseline Analysis for Controlled Singing Voice Deepfake Detection,” arXiv preprint arXiv:2406.02438 , 2024

    2024 arXiv

Pith tools

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