Pith. sign in

REVIEW 4 major objections 8 minor 34 references

Mitigating Over-Suppression in Speech Enhancement via Inference-Time Rethink-and-Refine Correction Module

T0 review · 4 major / 8 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Over-suppression in speech enhancement can be corrected at inference time by per-segment remixing of the noisy and enhanced signals, guided by ASR alignments and optimized against a composite quality-and-preservation objective, with no…

desk verdict Useful plug-in idea for fixing over-suppression at inference time, but the robustness objective R is unspecified and several tables look wrong; deserves a major-revision round, not acceptance as-is. read the letter →

arxiv 2608.07781 v1 pith:CHIWNQRD submitted 2026-08-07 eess.AS

Mike Qu , Yu-Wen Chen , Julia Hirschberg This is my paper · ORCID
classification eess.AS
keywords speechenhancementover-suppressioninference-timecorrectionconvexinterpolationnon-intrusivequalityassessmentASR-guidedsegmentationrethink-and-refine
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

Speech enhancement models sometimes over-suppress: they remove genuine speech cues together with the noise, and the output can be less intelligible and less useful for downstream tasks than the unprocessed recording. This paper claims that this failure can be repaired after the fact, without retraining the enhancer, by treating the noisy input and the enhanced output as two views of the same utterance and locally blending them back together. An ASR model aligns both signals into word- or phoneme-level intervals, and for each interval the module picks a mixing weight that maximizes a composite objective built from a non-intrusive quality predictor and a speech-preservation score. Because only the damaged regions are remixed toward the noisy original, the denoising benefit is kept where the enhancer worked. Across three enhancers and four datasets the paper reports higher PESQ, STOI, and subjective ratings, and lower ASR word error rates, which is what makes the approach matter: existing enhancers can be wrapped by the module and immediately produce better and more usable speech.

What carries the argument

The load-bearing object is the per-segment convex interpolation of Equation 1, $$\hat{s}^{\mathrm{corr}}_i(t,\lambda_i) = \lambda_i \hat{s}_i(t) + (1-\lambda_i)x_i(t)$$ with $\lambda_i \in [0,1]$, which turns over-suppression repair into a per-segment blending problem. ASR-guided segmentation (Charsiu for phoneme alignments, Whisper-timestamped with DTW for word alignments) supplies the paired intervals; the weights are chosen by gradient descent on the composite objective $L(\lambda) = \gamma Q(s^{\mathrm{corr}}) + \delta R(s^{\mathrm{corr}})$, with $Q$ the SCOREQ non-intrusive quality predictor and $R$ the Whisper-embedding L2 similarity. Reconciliation runs the optimization on both the noisy-derived and enhanced-derived segment boundaries and keeps the higher-scoring reconstruction, and a final fallback retains the original noisy or enhanced signal if either scores higher. This machinery is what allows a black-box enhancer to be treated as a first-pass generator whose unreliable regions can be re-estimated, making a training-free correction stage possible.

What would settle it

Run the module on a deliberately damaged utterance in which an enhancer over-suppresses a phoneme by time-reversing that segment's waveform; a time-reversed segment and the noisy segment span only their own samples, so no $\lambda \in [0,1]$ can reproduce the time-forward clean segment, and if PESQ/STOI of the corrected output does not beat the better of noisy and enhanced on that segment, the convex-hull repair assumption is falsified. A lighter check: inject a pure tone into a segment at a frequency absent from both inputs and verify that optimal interpolation leaves the artifact in the output.

Watch

Extended reading notes

Core claim

The central claim is that the destructive error of over-suppression is local, detectable, and correctable from observable signals alone. For each ASR-aligned segment there exists a weight $\lambda_i \in [0,1]$ such that replacing the enhanced segment with the convex combination $\lambda_i \hat{s}_i(t) + (1-\lambda_i)x_i(t)$ improves perceptual quality relative to the enhancer's output while preserving speech content. The module never modifies the enhancer; it optimizes the whole vector of segment weights against a composite objective, compares the corrected waveform, the enhanced waveform, and the noisy waveform under the same objective, and keeps the highest-scoring one. Across CMGAN, SEMamba, and SGMSE+ on URGENT 2024/2025, VCTK-DEMAND, and MSP-PODCAST/Audioset, the paper reports consistent gains in PESQ, STOI, SCOREQ MOS, and aesthetic scores, plus reductions in ASR word error rate and speaker-embedding distance, and a listening test rating corrected audio above the enhanced baseline.

Load-bearing premise

The repair is only as good as the two signals being blended: every over-suppressed segment is assumed to be fixable by some straight-line mix of the enhanced segment and the original noisy segment, so any enhancer error that neither input contains (phase reversals, pitch shifts, inserted artifacts) falls outside what the module can restore.

Editorial extensions

If this is right

  • Pretrained speech enhancers can be upgraded at inference time: wrapping any existing model with the module adds a second pass that repairs over-suppressed segments, so deployments do not need retraining or new paired data.
  • Downstream speech tasks benefit directly: corrected audio lowers ASR word error rates and speaker-verification distances relative to enhanced audio, so transcription and speaker-identification systems receive more faithful inputs.
  • Because the optimizer works per segment, well-denoised regions are left mostly enhanced while only unreliable intervals are blended back toward the noisy signal, avoiding the uniform noise reintroduction of global mixing approaches.
  • The fallback rule (keep the better of noisy, enhanced, or corrected under the objective) makes the correction conservative: in principle it should never degrade the chosen baseline on that objective.
  • The framework separates enhancement from correction, mirroring the rethink-and-refine pattern from language-model decoding, which suggests speech-processing pipelines can be built with independent first-pass and repair stages.

Reading between the lines

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

  • The convex-hull repair bound means the method's ceiling is set by the noisy and enhanced inputs: if an enhancer introduces distortion not representable as a blend of the two (phase flips, pitch shifts, inserted hallucinated tones), no $\lambda$ can reconstruct the clean speech; a richer repair family such as spectral masking or generative inpainting would be needed.
  • The robustness term $R$ is Whisper-embedding similarity, which ties correction to what ASR attends to; swapping $R$ for an emotion- or speaker-oriented embedding would likely reshape which segments get repaired, meaning the module is a general scaffolding whose behavior is set by the chosen objective.
  • The per-segment weights are optimized by gradient descent at test time, so latency scales with the number of segments and optimizer steps; a learned predictor that maps segment features to mixing weights could amortize the correction for real-time use, though the paper leaves efficiency to future work.
  • Reconciliation currently chooses between two segmentation views (noisy-derived and enhanced-derived); extending to several aligners or a minimum-cost boundary merge might reduce sensitivity to segmentation errors, an avenue the paper does not explore.
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 / 8 minor

Summary. The paper proposes an inference-time "rethink-and-refine" correction module for speech enhancement. Given a noisy signal and the output of a pretrained SE model, the method uses ASR-based word/phoneme alignment to segment both signals, then optimizes per-segment convex mixing weights so that the reconstructed waveform maximizes a composite objective combining a non-intrusive quality score Q and a speech-preservation robustness score R. The authors claim that this plug-in module improves PESQ, STOI, and downstream ASR/speaker-verification metrics over the baseline enhancer across three SE models and several datasets, while requiring no retraining. A subjective listening study on a subset of URGENT 2024 is also reported. The central idea is clear and the experiments cover a reasonable range of conditions, but the definition of the robustness term R is incomplete, and one of the headline evaluation metrics (SCOREQ) is also the optimization objective, which weakens the evidence as currently presented.

Significance. If the claimed gains are real and reproducible, the module would be a practically attractive post-processor: it treats the base SE model as a black box, requires no training, and could be attached to arbitrary enhancers to reduce over-suppression artifacts. The paper also evaluates downstream WER, speaker embedding distance, jitter/shimmer, and a small subjective test, which is broader than typical SE evaluations. The main limitation is that the empirical claim is currently conditional on an underspecified objective function: the R term in Eq. (2) is not defined as a computable function, and the SCOREQ metric appears both inside the optimized objective and as a reported result, creating a selection effect. The paper's value will be clear once the objective is fully specified and the independent metrics are separated from the optimized ones.

major comments (4)
  1. [Section 2.2, Eq. (2)] The robustness term R(·) is not defined precisely enough for the experiments to be reproducible. The text says only that R measures "Whisper embedding L2 similarity," and Section 3.1 repeats that Whisper-base is used "to compute a L2 distance-based robustness objective," but the reference signal for the L2 distance is never stated. If the reference is the clean speech, the method requires ground-truth clean audio at inference time, contradicting the abstract's claim that it operates "entirely in the inference stage" and Section 1's claim of "without retraining." If the reference is the noisy input, the optimizer is rewarded for pulling corrected segments toward the noisy signal, which would counteract noise reduction. If the reference is the enhanced signal itself, the term becomes vacuous. Because all reported gains in Tables 1-3 are produced by optimizing Eq. (2), the central experimental claim is not reproducible as written.
  2. [Section 2.2 and Section 3.2, Tables 1-2] There is a circularity between the optimization objective and the evaluation protocol. SCOREQ appears as the Q(·) term in Eq. (2), and SCOREQ MOS is also reported as a headline metric in Tables 1 and 2. The SCOREQ improvements are therefore partly selected by the optimizer rather than independently measured; they should not be cited as evidence of improved perceptual quality. In addition, the R term uses Whisper-base embeddings while WER is computed with Whisper-base, so the downstream WER gains are not fully independent of the objective either. The manuscript should report the optimized metric separately from independently evaluated metrics, or should replace SCOREQ in the reported tables with a held-out non-intrusive metric that does not appear in Eq. (2).
  3. [Section 4.1, Table 1 and Section 4.2, Table 2] The stated contribution of "consistent gains in intrusive metrics" is not fully supported by the tables. On VCTK-DEMAND in Table 1, PESQ drops from 3.559 (enhanced) to 3.415 (corrected), and in Table 2, SGMSE AB-Aes CE drops from 4.084 to 4.050 while PQ drops from 5.716 to 5.554. The text acknowledges "most cases" for some metrics, but the abstract and contribution (2) claim consistent gains. Moreover, the tables are labeled "mean±std" but only point estimates are shown, so the reader cannot assess whether the improvements are statistically meaningful; many of the reported differences are likely within one standard deviation. Please either qualify the central claim to match the actual per-metric outcomes or provide per-utterance statistical tests with standard deviations and significance levels.
  4. [Section 2.1, Eq. (1)] The repair mechanism is bounded by the convex hull of the noisy and enhanced segments. If the enhancer introduces phase, spectral, or temporal distortions that are not present in either input, Eq. (1) cannot undo them, and the optimizer can only choose how much of each input to keep. The paper should explicitly acknowledge this limitation and discuss what kinds of over-suppression artifacts are actually correctable by this interpolation. This does not invalidate the approach, but it narrows the claim that the module can be applied to "any off-the-shelf SE model" for arbitrary corruption types.
minor comments (8)
  1. [Fig. 1 and Section 2.2] The figure uses θ and α for the mixing coefficients while Section 2.2 uses λ; the notation should be unified for clarity.
  2. [Fig. 1 and Eq. (2)] The figure shows a minimization of -γQ - δR while Eq. (2) defines maximization of γQ + δR; the sign convention should be made consistent in both places.
  3. [Tables 1-3] The table headers say "mean±std" but the displayed values contain no standard deviations; either include the standard deviations or change the header labels.
  4. [Table 3, URGENT 2025 row] The enhanced WER of 226.4% is much larger than the noisy WER of 93.3%; since WER can exceed 100% when insertions dominate, the paper should explain this behavior and clarify whether the correction module mainly reduces insertions.
  5. [Table 3, URGENT 2024 row] The enhanced row reports jitter as 0.84×10⁻² while the clean row reports 2.36×10⁻²; this is surprising and should be verified, since enhanced speech is expected to have jitter closer to, or larger than, clean speech.
  6. [Section 3.2, subjective listening test] Only mean ratings are given for the listening test; the paper should report the number of utterances and annotators per condition and a more standard inter-rater reliability measure, since the reported 54.77% agreement is difficult to interpret.
  7. [Section 5] The method optimizes K mixing weights per utterance and evaluates the full reconstructed waveform multiple times, but no runtime or complexity analysis is provided. A brief wall-clock comparison to the base SE model would help assess practical usability.
  8. [Section 1] The prior global-mixing methods [6,7] are discussed qualitatively but no direct comparison to them is presented; adding such a comparison would strengthen the claim that per-segment optimization is beneficial.

Circularity Check

1 steps flagged · score 4.0 of 10

SCOREQ is optimized at inference and then reported as an evaluation gain; independent PESQ/STOI and listening results keep the central claim non-circular.

  1. self definitional [Section 2.2 (Eq. 2) and Section 3.2 (evaluation metrics)]
    "LetQ(·)denote a non-intrusive quality predictor (e.g., SCOREQ) ... We also include the SCOREQ MOS score [17], which was used to guide the weight selection."

    The segment weights are chosen by maximizing L(λ)=γQ(s_corr)+δR(s_corr), with Q instantiated as SCOREQ. The corrected waveform is therefore, by construction, the waveform that maximizes (among the allowed convex mixes) the SCOREQ component of the objective. Reporting the SCOREQ MOS of the corrected signal in Table 1 as evidence of improvement is a self-referential evaluation: the same function used to select the weights is also used as the headline metric. The SCOREQ gain is partly selected for rather than independently measured. This does not invalidate the PESQ, STOI, or listening-test results, which are not directly optimized, so the circularity is localized to the SCOREQ column.

full rationale

The central derivation is not circular as a whole. The correction module optimizes per-segment interpolation weights (Eqs. 1-3) at inference time, and the main claims (PESQ/STOI gains, WER reduction, vocal-trait preservation) are evaluated with metrics that are not directly used in the objective. The one clear circular element is the SCOREQ column in Tables 1 and 2: SCOREQ appears in the objective L (Eq. 2), so the reported SCOREQ improvement is a self-referential consequence of the optimization procedure rather than an independent measurement. That partial circularity does not extend to PESQ, STOI, human listening, SpeakerNet, jitter, or shimmer, which give the central claim independent content. Two additional concerns are flagged but are not fully demonstrated circularity: (i) the robustness term R is specified only as 'Whisper embedding L2 similarity' with no reference signal, leaving the inference-only claim unverifiable; (ii) because Whisper-base is used both inside the objective and for the WER evaluation, the WER result could share the optimization signal, but the absence of a defined reference prevents an equation-level proof of that overlap. Self-citations [7] and [29] are contextual (prior interpolation approach and subjective-preference evidence) and are not load-bearing in the derivation.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

No new physical or conceptual entities are postulated. The method relies on existing ASR and quality-assessment models and on manipulation of existing signals.

free parameters (2)
  • Segment-wise mixing weights lambda_i = Optimized per utterance, final values not reported
    Eq. 3 treats lambda_i as inference-time decision variables; the correction is tuned per segment, and the fitted values are not listed.
  • Objective trade-off weights gamma, delta = 0.5, 0.5
    Set by hand in Section 2.2 with no ablation; the central claim depends on this balance between quality and speech preservation.
assumptions (3)
  • domain assumption ASR timestamps from Charsiu or Whisper-timestamped provide valid word or phoneme intervals on both noisy and enhanced speech.
    Section 2.1 builds all segmentation from these alignments; if timestamps are wrong, as may happen when enhanced WER exceeds 200% on URGENT2025, the intervals may not localize over-suppression.
  • domain assumption The composite objective L(lambda) = gamma*Q + delta*R correctly ranks candidate corrections on unseen noise conditions.
    Eq. 2 is the only guide for choosing lambda; its validity is assumed, though partially supported by independent PESQ, STOI, and human listening results.
  • domain assumption Clean speech in an over-suppressed segment can be approximated by a convex combination of the noisy and enhanced segments.
    Eq. 1 restricts all repairs to linear interpolation; it cannot recover speech if the enhancer introduced distortions outside the convex hull of the two inputs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mitigating Over-Suppression in Speech Enhancement via Inference-Time Rethink-and-Refine Correction Module." pith.science (2026). https://pith.science/paper/CHIWNQRD

@misc{pith2026260807781,
  author       = {Pith},
  title        = {Pith review of: Mitigating Over-Suppression in Speech Enhancement via Inference-Time Rethink-and-Refine Correction Module},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CHIWNQRD}},
  note         = {Machine review of arXiv:2608.07781}
}
read the original abstract

We present a rethink-and-refine correction module that addresses over-suppression, a common failure mode of speech enhancement (SE) models, where speech cues are suppressed alongside noise. Our method operates entirely in the inference stage without additional training, allowing seamless integration with diverse SE models. Given noisy and enhanced signals, we obtain word- or phoneme-level alignments using an automatic speech recognition model and identify intervals where enhancement is unreliable. These intervals are then selectively remixed through convex interpolation, with per-segment weights optimized to maximize a composite objective balancing perceptual quality and speech preservation. Experiments on the URGENT 2024 and 2025, VCTK-DEMAND, and MSP-PODCAST datasets show consistent improvements in perceptual quality, intelligibility, and downstream performance compared to conventional SE alone, demonstrating the benefit of rethink-and-refine framework for robust speech processing.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 30 canonical work pages

  1. [1]

    Mitigating Over-Suppression in Speech Enhancement via Inference-Time Rethink-and-Refine Correction Module

    INTRODUCTION Speech enhancement (SE) models aim to suppress noise while preserving speech quality and intelligibility. Recent deep learning-based approaches [1, 2, 3] are typically trained on tuples of noisy and clean waveforms, with the objective of manipulating the noisy signal to make it as “similar” to the clean signal as possible [4, 5]. When test co...

    work page Pith review arXiv 2026

  2. [2]

    METHODOLOGY We propose a correction module for SE, inspired by the rethink-and-refine framework, which identifies parts of the input speech where the initial SE over-suppresses the signal and selectively restores the information (Fig. 1). First, we segment both the noisy and output of pretrained SE model (i.e., enhanced signals) into word- or phoneme-leve...

    work page

  3. [3]

    Models and datasets We evaluate our approach using three state-of-the-art SE models, including CMGAN [1], SEMamba [2], and SGMSE [3]

    EXPERIMENTAL SETUP 3.1. Models and datasets We evaluate our approach using three state-of-the-art SE models, including CMGAN [1], SEMamba [2], and SGMSE [3]. For fair comparison, all models use official V oiceBank- DEMAND pre-trained checkpoints: the standard CMGAN checkpoint,SEMamba advanced.pth, and SGMSE+. To guide weight selection, we use Whisper-base...

    work page 2024

  4. [4]

    RESULTS 4.1. Performance across datasets We evaluate SEMamba [2] both with and without the pro- posed output-level correction module across multiple bench- mark datasets to examine whether our approach can reliably improve a strong SE under unseen noise conditions (Ta- ble 1). On every dataset, our method yields higher STOI and SCOREQ MOS than the enhance...

    work page arXiv 2025

  5. [5]

    CONCLUSION We introduce a correction module applied to existing SE mod- els to mitigate destructive over-suppression while preserving denoising benefits under unseen noise conditions. Results show that our method is compatible with various SE archi- tectures and generally improves performance across datasets in both SA metrics and downstream evaluations c...

    work page

  6. [6]

    CM- GAN: Conformer-based metric GAN for speech en- hancement,

    Ruizhe Cao, Sherif Abdulatif, and Bin Yang, “CM- GAN: Conformer-based metric GAN for speech en- hancement,” inProc. INTERSPEECH, 2022, pp. 936– 940

    work page 2022

  7. [7]

    An investigation of incorporating mamba for speech enhancement,

    Rong Chao, Wen-Huang Cheng, Moreno La Quatra, Sabato Marco Siniscalchi, Chao-Han Huck Yang, Szu- Wei Fu, and Yu Tsao, “An investigation of incorporating mamba for speech enhancement,” inProc. SLT, 2024, pp. 302–308

    work page 2024

  8. [8]

    Speech enhance- ment and dereverberation with diffusion-based genera- tive models,

    Julius Richter, Simon Welker, Jean-Marie Lemercier, Bunlong Lay, and Timo Gerkmann, “Speech enhance- ment and dereverberation with diffusion-based genera- tive models,”IEEE/ACM Trans. Audio, Speech, Lang. Process., vol. 31, pp. 2351–2364, 2023

    work page 2023

Show all 34 references
  1. [9]

    Dense CNN with self-attention for time-domain speech enhance- ment,

    Ashutosh Pandey and DeLiang Wang, “Dense CNN with self-attention for time-domain speech enhance- ment,”IEEE/ACM Trans. Audio, Speech, Lang. Pro- cess., vol. 29, pp. 1270–1279, 2021

    2021

  2. [10]

    MP-SENet: A speech enhancement model with parallel denoising of magnitude and phase spectra,

    Ye-Xin Lu, Yang Ai, and Zhen-Hua Ling, “MP-SENet: A speech enhancement model with parallel denoising of magnitude and phase spectra,” inProc. INTERSPEECH, 2023, pp. 3834–3838

    2023

  3. [11]

    How bad are artifacts?: Analyzing the im- pact of speech enhancement errors on ASR,

    Kazuma Iwamoto, Tsubasa Ochiai, Marc Delcroix, Rin- taro Ikeshita, Hiroshi Sato, Shoko Araki, and Shigeru Katagiri, “How bad are artifacts?: Analyzing the im- pact of speech enhancement errors on ASR,” inProc. INTERSPEECH, 2022, pp. 5418–5422

    2022

  4. [12]

    Noise robust speech emotion recognition with signal-to-noise ratio adapting speech enhancement,

    Yu-Wen Chen, Julia Hirschberg, and Yu Tsao, “Noise robust speech emotion recognition with signal-to-noise ratio adapting speech enhancement,”arXiv preprint arXiv:2309.01164, 2023

    2023 arXiv

  5. [13]

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

    A.W. Rix, J.G. Beerends, M.P. Hollier, and A.P. Hek- stra, “Perceptual evaluation of speech quality (pesq)-a new method for speech quality assessment of telephone networks and codecs,” inProc. ICASSP, 2001, vol. 2, pp. 749–752

    2001

  6. [14]

    An algorithm for intelligibility prediction of time- frequency weighted noisy speech,

    C. H. Taal, R. C. Hendriks, R. Heusdens, and J. Jensen, “An algorithm for intelligibility prediction of time- frequency weighted noisy speech,”IEEE/ACM Trans. Audio, Speech, Lang. Process., vol. 19, no. 7, pp. 2125– 2136, 2011

    2011

  7. [15]

    MetricGAN- OKD: Multi-metric optimization of MetricGAN via on- line knowledge distillation for speech enhancement,

    Wooseok Shin, Byung Hoon Lee, Jin Sob Kim, Hyun Joon Park, and Sung Won Han, “MetricGAN- OKD: Multi-metric optimization of MetricGAN via on- line knowledge distillation for speech enhancement,” in Proc. ICML, 2023, vol. 202, pp. 31521–31538

    2023

  8. [16]

    Multi-CMGAN+/+: Leveraging multi- objective speech quality metric prediction for speech en- hancement,

    George Close, William Ravenscroft, Thomas Hain, and Stefan Goetze, “Multi-CMGAN+/+: Leveraging multi- objective speech quality metric prediction for speech en- hancement,” inProc. ICASSP, 2024, pp. 351–355

    2024

  9. [17]

    Reflexion: Lan- guage agents with verbal reinforcement learning,

    Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao, “Reflexion: Lan- guage agents with verbal reinforcement learning,” in Proc. NeurIPS, 2023, vol. 36, pp. 8634–8652

    2023

  10. [18]

    Self- refine: Iterative refinement with self-feedback,

    Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, et al., “Self- refine: Iterative refinement with self-feedback,” inProc. NeurIPS, 2023, vol. 36, pp. 46534–46594

    2023

  11. [19]

    Phone- to-audio alignment without text: A semi-supervised ap- proach,

    Jian Zhu, Cong Zhang, and David Jurgens, “Phone- to-audio alignment without text: A semi-supervised ap- proach,” inProc. ICASSP, 2022, pp. 8167–8171

    2022

  12. [20]

    Robust speech recognition via large-scale weak supervision,

    Alec Radford, Jong Wook Kim, Tao Xu, Greg Brock- man, Christine McLeavey, and Ilya Sutskever, “Robust speech recognition via large-scale weak supervision,” in Proc. ICML, 2023, pp. 28492–28518

    2023

  13. [21]

    Computing and visualizing dynamic time warping alignments in r: The dtw package,

    Toni Giorgino, “Computing and visualizing dynamic time warping alignments in r: The dtw package,”Jour- nal of Statistical Software, vol. 31, no. 7, 2009

    2009

  14. [22]

    SCOREQ: Speech quality assessment with contrastive regression,

    Alessandro Ragano, Jan Skoglund, and Andrew Hines, “SCOREQ: Speech quality assessment with contrastive regression,” inProc. NeurIPS, 2024, vol. 37, pp. 105702–105729

    2024

  15. [23]

    URGENT challenge: Universal- ity, robustness, and generalizability for speech enhance- ment,

    Wangyou Zhang, Robin Scheibler, Kohei Saijo, Samuele Cornell, Chenda Li, Zhaoheng Ni, Jan Pirkl- bauer, Marvin Sach, Shinji Watanabe, Tim Fingscheidt, and Yanmin Qian, “URGENT challenge: Universal- ity, robustness, and generalizability for speech enhance- ment,” inProc. INTERS...

    2024

  16. [24]

    Interspeech 2025 URGENT speech enhancement challenge,

    Kohei Saijo, Wangyou Zhang, Samuele Cornell, Robin Scheibler, Chenda Li, Zhaoheng Ni, Anurag Kumar, Marvin Sach, Yihui Fu, Wei Wang, Tim Fingscheidt, and Shinji Watanabe, “Interspeech 2025 URGENT speech enhancement challenge,” inProc. INTER- SPEECH, 2025, pp. 858–862

    2025

  17. [25]

    Investigating RNN- based speech enhancement methods for noise-robust Text-to-Speech,

    Cassia Valentini-Botinhao, Xiaofei Wang, Shingo Takaki, and Junichi Yamagishi, “Investigating RNN- based speech enhancement methods for noise-robust Text-to-Speech,” inProc. SSW, 2016

    2016

  18. [26]

    Building naturalistic emotionally balanced speech corpus by retrieving emo- tional speech from existing podcast recordings,

    Reza Lotfian and Carlos Busso, “Building naturalistic emotionally balanced speech corpus by retrieving emo- tional speech from existing podcast recordings,”IEEE Transactions on Affective Computing, vol. 10, no. 4, pp. 471–483, 2017

    2017

  19. [27]

    Audio set: An on- tology and human-labeled dataset for audio events,

    Jort F. Gemmeke, Daniel P. W. Ellis, Dylan Freedman, Aren Jansen, Wade Lawrence, R. Channing Moore, Manoj Plakal, and Marvin Ritter, “Audio set: An on- tology and human-labeled dataset for audio events,” in Proc. ICASSP, 2017, pp. 776–780

    2017

  20. [28]

    Meta audiobox aesthetics: Unified automatic quality assessment for speech, music, and sound,

    Andros Tjandra, Yi-Chiao Wu, Baishan Guo, John Hoff- man, Brian Ellis, Apoorv Vyas, Bowen Shi, Sanyuan Chen, Matt Le, Nick Zacharov, Carleigh Wood, Ann Lee, and Wei-Ning Hsu, “Meta audiobox aesthetics: Unified automatic quality assessment for speech, music, and sound,”arXiv pr...

    2025 arXiv

  21. [29]

    SpeakerNet: 1D depth-wise separable convolutional network for text-independent speaker recognition and verification,

    Nithin Rao Koluguri, Jason Li, Vitaly Lavrukhin, and Boris Ginsburg, “SpeakerNet: 1D depth-wise separable convolutional network for text-independent speaker recognition and verification,”arXiv preprint arXiv:2010.12653, 2020

    2010 arXiv

  22. [30]

    V ocal acoustic analysis–jitter, shimmer and hnr param- eters,

    Jo ˜ao Paulo Teixeira, Carla Oliveira, and Carla Lopes, “V ocal acoustic analysis–jitter, shimmer and hnr param- eters,”Procedia technology, vol. 9, pp. 1112–1122, 2013

    2013

  23. [31]

    The psychologist as an interlocutor in autism spectrum disorder assessment: Insights from a study of spontaneous prosody,

    Daniel Bone, Chi-Chun Lee, Matthew P Black, Mar- ian E Williams, Sungbok Lee, Pat Levitt, and Shrikanth Narayanan, “The psychologist as an interlocutor in autism spectrum disorder assessment: Insights from a study of spontaneous prosody,”Journal of Speech, Lan- guage, and Hear...

    2014

  24. [32]

    In- troducing Parselmouth: A Python interface to Praat,

    Yannick Jadoul, Bill Thompson, and Bart de Boer, “In- troducing Parselmouth: A Python interface to Praat,” Journal of Phonetics, vol. 71, pp. 1–15, 2018

    2018

  25. [33]

    Too good to be true: A study on modern automatic speech recognition for the evaluation of speech enhancement,

    Danilo de Oliveira, Tal Peer, and Timo Gerkmann, “Too good to be true: A study on modern automatic speech recognition for the evaluation of speech enhancement,” arXiv preprint arXiv:2605.12107, 2026

    2026 arXiv

  26. [34]

    InQSS: a speech intelligi- bility and quality assessment model using a multi-task learning network,

    Yu-Wen Chen and Yu Tsao, “InQSS: a speech intelligi- bility and quality assessment model using a multi-task learning network,” inProc. INTERSPEECH, 2022, pp. 3088–3092

    2022

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.