Time-Graph Frequency Representation with Singular Value Decomposition for Neural Speech Enhancement

Meng Ge; Qiquan Zhang; Tianrui Wang; Tingting Wang; Zhen Yang; Zirui Ge

REVIEW 4 major objections 6 minor 34 references

Time-Graph Frequency Representation with Singular Value Decomposition for Neural Speech Enhancement

T0 review · 4 major / 6 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 A real-valued graph-frequency transform built from singular vectors lets one mask stream carry amplitude and phase, outperforming STFT and complex GFT-EVD front-ends for neural speech enhancement.

desk verdict Useful front-end trick with consistent gains, but the evaluation needs a validation-based k and significance testing before the strong claim is taken as proven. read the letter →

arxiv 2412.16823 v2 pith:WC5JTUHS submitted 2024-12-22 eess.AS

Tingting Wang , Tianrui Wang , Meng Ge , Qiquan Zhang , Zirui Ge , Zhen Yang This is my paper · ORCID

classification eess.AS

keywords speechenhancementgraphFouriertransformsingularvaluedecompositionreal-valuedtime-graphrepresentationmonauralspectralmappingdeeplearning

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that a real-valued 'time-graph' spectrogram, produced by a graph Fourier transform whose basis comes from the singular value decomposition of a 0-1 adjacency matrix, lets a single mask-estimating network represent both amplitude and phase at once. The point is to avoid the dual-stream magnitude/phase or real/imaginary designs, which the authors argue suffer from misalignment between their two branches. If the claim holds, any mask-based enhancement network could swap its STFT front-end for GFT-SVD and gain objective perceptual quality and intelligibility without adding parameters. The authors test the idea with seven backbone networks on two benchmarks and report that GFT-SVD consistently improves PESQ, STOI, and SI-SDR over the same backbones with STFT, and over the earlier GFT-EVD baseline with a UNet.

What carries the argument

The load-bearing object is the 0-1 adjacency matrix $A_k$, defined on each speech frame by setting $A_k(i,j)=1$ when speech sample $j$ is one of the $k$ neighbours of sample $i$; the paper uses $k=3$ after a small empirical sweep. GFT-SVD decomposes $A_k = \Psi \Lambda \Gamma$ and uses the real unitary matrix $\Psi$ as the graph-Fourier basis, mapping a frame $y$ to $Y_G = \Psi y$ and reconstructing with $s = \Psi^{-1} \hat{S}_G$. Because $\Psi$ is real and orthogonal, the representation and the whole mask-then-inverse pipeline stay real-valued, which is what removes the need for a dual-stream complex network.

What would settle it

To test the central claim, choose $k$ using only a validation split (or fix $k=3$), retrain on a fresh training set, and evaluate on a noise corpus the model has not seen; if GFT-SVD no longer beats the STFT front-end on PESQ and SI-SDR, the reported advantage is tied to the tuned setting. A sharper control is to replace $A_k$ with a random 0-1 matrix of the same sparsity: if the same gains appear, the speech-specific adjacency is not what carries the result.

Watch

Extended reading notes

Core claim

The paper's core discovery is that the singular value decomposition of the non-symmetric 0-1 adjacency matrix $A_k$ yields a real orthogonal graph-Fourier basis $\Psi$, so $Y_G = \Psi y$ is a real-valued, invertible graph spectrum that carries amplitude and phase information together. Because the encoder, mask product, and decoder are all real operations, the pipeline never has to estimate phase separately or align two network branches. In the reported experiments, this real-valued time-graph representation outperforms the complex STFT representation on the same backbone networks and outperforms the complex eigenvector-based GFT-EVD when both are paired with a UNet, on wideband and narrowband PESQ, SI-SDR, and STOI.

Load-bearing premise

The argument depends on the idea that connecting each speech sample to its three or five nearest neighbours creates a graph whose left singular vectors are a meaningful speech-frequency basis for denoising; if that graph structure is arbitrary, the gains could be specific to the tuned neighbourhood size and the two test corpora.

Editorial extensions

If this is right

Any mask-based enhancement network can use GFT-SVD as a drop-in front-end; in the paper's tests all seven backbones improve on most objective metrics relative to their STFT versions without changing architecture or parameter count.
The real-valued representation removes the phase-estimation branch, so the estimated waveform comes from a single real matrix multiply and the pipeline avoids phase wrapping and amplitude-phase alignment errors.
On the DNS-2020 no-reverb test set, DPCRN-G, DCRN-G, and UNet-G each gain more than 1 dB SI-SDR over their STFT counterparts and over the GFT-EVD UNet baseline.
GFT-SVD adds essentially no parameters and only a small increase in MACs, while the reported real-time factor is lower than the corresponding STFT model, so the improvement is not bought with extra model capacity.
Because the transform is invertible and real-valued, it can replace the STFT front-end in masked-based single-channel enhancement systems generally, not just in the backbones tested here.

Reading between the lines

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

The same front-end substitution could likely transfer to other single-channel mask-based tasks such as speaker separation or denoising for ASR, because the transform itself is agnostic to the mask estimator; the paper only reports speech-enhancement results.
The strong dependence on k (best at k=3 or 5, degrading as k grows) suggests the singular vectors are acting like a fixed local difference operator; a direct comparison with a graph Laplacian or a learned adjacency would show whether the specific graph structure or just the real-valued property is responsible.
Because the whole pipeline is real-valued, the approach could let complex-valued architectures run as real networks with fewer parameters; the paper reports lower real-time factors but does not retrain a complex-network variant to test this.

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.

Desk Editor's Note

read the letter

Short version: the paper shows a real-valued SVD-based graph transform front-end that gives consistent 0.1–0.3 PESQ and up to ~1.5 dB SI-SDR gains over STFT across eight backbones and two datasets. That's a genuine, modest contribution. But the evaluation has a few fixable weaknesses that prevent me from fully trusting the headline claim.

What's new: The GFT-SVD is a decomposition substitution on the 0-1 adjacency matrix: use left singular vectors as the basis instead of eigenvectors. This yields a real-valued analysis/synthesis pair, which is a neat way to avoid complex-valued spectrograms and their two-stream decoupling. The experimental sweep is thorough for a front-end proposal, and the fact that gains appear in nearly every backbone gives the result some weight. Code is promised.

Soft spots, in order of severity: - k is selected on the test data. Section IV-C uses the same test sets as Tables I and II to pick k=3 or 5, so the reported gains are max-over-k numbers. This is a real inflation risk, though the consistency across eight backbones makes pure overfitting unlikely. A validation-based k plus a sensitivity statement would fix it. - No significance tests or error bars. The gains are small enough that a paired test would be appropriate. - For MTFAA and BSRNN, the GFT-SVD version removes the band merging/splitting process, so those rows are not architecture-matched comparisons. That's a confound for two of the eight backbones. - The GFT-EVD baseline is only run with UNet (G-UNet). So the abstract's claim of outperforming GFT-EVD is actually a claim against one backbone, not a broad comparison. - The 'graph frequency' interpretation is loose. SVD singular vectors of a k-neighbor adjacency matrix aren't a spectral basis in the standard graph signal processing sense. The paper would be stronger if it framed this as a data-dependent orthonormal transform rather than a GFT.

What holds up: The transform is linear, invertible, and parameter-free except for k. The central empirical pattern is consistent and the protocol is otherwise reasonable. None of the weaknesses are load-bearing; they're all addressable with extra experiments.

I'd send this to peer review. A good referee should push for validation-based k selection, significance testing, matched backbones for all rows, and a wider GFT-EVD comparison. The paper is useful for speech enhancement people who want a cheap front-end replacement that avoids complex-valued modeling.

Referee Report

4 major / 6 minor

Summary. The paper proposes GFT-SVD, a real-valued time-graph representation for monaural speech enhancement. The transform projects noisy waveform frames onto the left singular vectors of a 0-1 graph adjacency matrix A_k that encodes k-nearest-neighbor dependencies among speech samples. The resulting real-valued coefficients replace the complex STFT spectrogram, and a DNN mask estimator is trained with SI-SDR loss. The authors evaluate the combination of GFT-SVD with eight backbone networks on the DNS-2020 and VCTK+DEMAND datasets, reporting consistent improvements over the same backbones with STFT and over a GFT-EVD-based graph UNet. Code is released.

Significance. If the reported gains are reliable, the paper offers a simple, real-valued front-end that could replace complex STFT features in a wide class of neural speech enhancement systems, with no change to the network architecture. The cross-backbone consistency (eight architectures, two datasets) is a strength, as is the public release of source code. However, the load-bearing claims depend on the selection of the graph neighbor count k and on the statistical validity of the reported differences, both of which are currently under-supported. The theoretical grounding of the SVD-based graph Fourier transform is also thin, leaving the method's generality an open question.

major comments (4)

[§IV-C and Tables I-II] The parameter k is effectively selected on the test data. Figure 3 reports average WB-PESQ and SI-SDR as a function of k, and all comparison tables use A3. The same DNS-2020 no-reverb and VCTK+DEMAND test sets used for Tables I and II appear to drive this selection. Thus the reported GFT-SVD gains are maximum-over-k values, not the performance of a model whose hyperparameter was chosen before seeing the test data. To support the central claim that the real-valued time-graph representation provides a general front-end improvement, the authors should select k on the validation split (or via nested cross-validation) and then report the corresponding test results, or report results for all k values and show that the ordering is stable. Without this, the advantage over STFT/GFT-EVD may be inflated by test-set overfitting.
[Tables I-II and Fig. 3] The paper reports point estimates only, with no error bars, confidence intervals, or paired significance tests. Several reported gains are small; for example, MTFAA_G over MTFAA in WB-PESQ is 0.011 on DNS-2020 (2.707 vs 2.696), and DCRN_G over DCRN in STOI is 0.004. Given typical run-to-run variance in training neural speech enhancement models, such differences may not be statistically reliable. The claim of consistent improvement across backbones would be considerably stronger with multiple training runs and paired significance tests (e.g., Wilcoxon signed-rank or a paired t-test on per-utterance scores).
[§III-B and §II-A] The paper asserts, without proof or analysis, that the left singular vectors of the 0-1 adjacency matrix A_k form a valid graph Fourier basis and that this basis captures speech structure. No graph Laplacian or frequency interpretation is defined; the SVD basis is simply a data-independent orthonormal basis determined by the binary matrix A_k. The only structural claim is that k=3 or 5 is best, but this is established empirically on test data. The authors should either provide a theoretical argument for why SVD of A_k yields a meaningful frequency-like representation, or present an ablation that varies the basis (e.g., random orthonormal basis, PCA of the training set) to show the specific choice of A_k matters. This concern is load-bearing because the claimed advantage of GFT-SVD over STFT is attributed to the graph structure, not merely to using a learned or fixed real-valued transform.
[§IV-A] The fairness of the STFT baseline dimensionality is unclear. The paper states that 512-dimension graph features are used for GFT-SVD and that "the 512-dimension features after STFT are inputted into the baseline." A 25 ms window at 16 kHz with a 512-point FFT yields 257 complex STFT coefficients, or 514 real values if real and imaginary parts are concatenated, or 257 magnitude values if only magnitude is used. The paper does not specify which STFT feature configuration is used. If STFT baselines use 512 real values (e.g., magnitude only) while GFT-SVD uses 512 real values, the comparison may favor GFT-SVD because it carries both amplitude and phase information in the same dimensionality. The authors should specify the exact STFT feature representation and, if the information content differs, discuss the implications for the comparison.

minor comments (6)

[Abstract] The abstract contains a typo: "neutral speech enhancement" should be "neural speech enhancement." A similar typo appears in the Introduction ("neutral speech enhancement").
[Eq. (8)] The SI-SDR loss formula is typeset awkwardly; the numerator and denominator are not clearly separated. Please rewrite the equation with explicit fraction formatting so that the loss definition is unambiguous.
[Table III] The last column header is "RFT" but the text refers to the real-time factor (RTF). Please correct the table header for consistency.
[§IV-A] Training details such as optimizer, learning rate, batch size, number of epochs, and loss weighting (if any) are not reported. Since code is released, these details would improve reproducibility and are expected for a journal submission.
[§II-A] The definition of A_k(i,j) says "dependency between the ith speech samples residing on vertex vi and the jth speech samples residing on vertex vj exists" but does not specify the distance metric or whether the graph is directed or undirected. Please clarify how k neighbors are chosen and why a directed 0-1 matrix is used rather than, for example, a weighted matrix with distance-dependent weights.
[Fig. 3] The figure is referenced but not shown in the manuscript text; please ensure the figure is included in the final version and that its axis labels and legend are legible. The text states "performance of all models will decrease with the increased value of k," but Figure 3 shows a peak at k=3 or 5; please make the description consistent with the actual plot.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: GFT-SVD is a fixed invertible transform and the empirical comparison is external; minor self-citations are not load-bearing, though k selection lacks a stated validation split.

full rationale

The central derivation is self-contained. Eq. (6) defines YG = Ψy with Ψ the left singular vectors of the 0-1 adjacency Ak; since Ψ is unitary, Eq. (7) gives an exact inverse, so the 'near-lossless' claim is a mathematical property, not an input assumed. The network is trained with the standard SI-SDR objective stated in Eq. (8), and all evaluation uses external corpora (DNS-2020, VCTK+DEMAND) and standard metrics (WB-PESQ, NB-PESQ, SI-SDR, STOI), so the reported superiority of GFT-SVD over STFT/GFT-EVD is an empirical result, not a consequence of the definitions. Self-citations [20] and [27] are present: [20] motivates the graph-adjacency view of speech, and [27] is cited for the SI-SDR loss, but neither carries the argument; the adjacency is a definition and the loss is written out explicitly, so they are not load-bearing. The one legitimate concern is Section IV-C: the choice k=3 ('A3') is reported as best in Fig. 3 and then used in Tables I and II, and the text does not state that this selection was made on a held-out validation split. If Fig. 3 used the same test sets as the comparison tables, the GFT-SVD gains are maximum-over-k values. That is a correctness/overfitting risk, but it is not circular: even at the selected k the comparison could in principle favor STFT/GFT-EVD, so the central claim is not forced by construction. Score 2 reflects only the minor, non-load-bearing self-citations.

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

The central claim rests mainly on the empirical choice of the graph construction and the SVD basis; there are no new physical entities. One free parameter (k) is tuned on test data, and three domain-specific modeling assumptions about the graph transform are unproven.

free parameters (1)

k (graph neighbor count) = 3 (or 5 for some settings)
Selected empirically in Fig. 3 to maximize average WB-PESQ and SI-SDR across models; the tables use this selected value.

assumptions (4)

domain assumption The 0-1 shift matrix A_k encodes meaningful dependency between neighboring speech samples.
Used in Section II-A to define the graph; no evidence is given that this specific adjacency is optimal or appropriate for speech.
ad hoc to paper Left singular vectors of A_k can serve as a graph Fourier basis.
Section III-B replaces the eigen-decomposition with SVD without justifying why singular vectors have a frequency interpretation or why this basis is better for masking.
domain assumption A real-valued mask in the GFT-SVD domain can reconstruct clean speech via the inverse transform.
Implicit in the pipeline (Eq. 4-5); invertibility alone does not guarantee effective denoising.
standard math SVD of a real matrix yields real orthogonal matrices.
Used to justify that GFT-SVD is real-valued and invertible.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Time-Graph Frequency Representation with Singular Value Decomposition for Neural Speech Enhancement." pith.science (2026). https://pith.science/paper/WC5JTUHS

@misc{pith2026241216823,
  author       = {Pith},
  title        = {Pith review of: Time-Graph Frequency Representation with Singular Value Decomposition for Neural Speech Enhancement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WC5JTUHS}},
  note         = {Machine review of arXiv:2412.16823}
}

read the original abstract

Time-frequency (T-F) domain methods for monaural speech enhancement have benefited from the success of deep learning. Recently, focus has been put on designing two-stream network models to predict amplitude mask and phase separately, or, coupling the amplitude and phase into Cartesian coordinates and constructing real and imaginary pairs. However, most methods suffer from the alignment modeling of amplitude and phase (real and imaginary pairs) in a two-stream network framework, which inevitably incurs performance restrictions. In this paper, we introduce a graph Fourier transform defined with the singular value decomposition (GFT-SVD), resulting in real-valued time-graph representation for neural speech enhancement. This real-valued representation-based GFT-SVD provides an ability to align the modeling of amplitude and phase, leading to avoiding recovering the target speech phase information. Our findings demonstrate the effects of real-valued time-graph representation based on GFT-SVD for neutral speech enhancement. The extensive speech enhancement experiments establish that the combination of GFT-SVD and DNN outperforms the combination of GFT with the eigenvector decomposition (GFT-EVD) and magnitude estimation UNet, and outperforms the short-time Fourier transform (STFT) and DNN, regarding objective intelligibility and perceptual quality. We release our source code at: https://github.com/Wangfighting0015/GFT\_project.

Figures

Figures reproduced from arXiv: 2412.16823 by the authors.

**Figure 2.** Example of the clean speech. The real and imaginary parts of clean [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 4.** (a), (b) The real and imaginary parts of the estimated mask in setting [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗

Figures from the paper (1 more)

**Figure 3.** Figure 3: The average WB-PESQ, SI-SDR of speech enhancement network [PITH_FULL_IMAGE:figures/full_fig_p004_3.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 23 canonical work pages

[1]

P. C. Loizou, Speech Enhancement: Theory and Practice, 2nd ed. Boca Raton, FL, USA: CRC Press, Inc., 2013

work page 2013
[2]

Supervised speech separation based on deep learning: An overview,

D. Wang and J. Chen, “Supervised speech separation based on deep learning: An overview,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 26, no. 10, pp. 1702–1726, 2018

2018
[3]

On MMSE-based estimation of amplitude and complex speech spectral coefficients under phase- uncertainty,

M. Krawczyk-Becker and T. Gerkmann, “On MMSE-based estimation of amplitude and complex speech spectral coefficients under phase- uncertainty,” IEEE/ACM Trans. Audio, Speech, and Lang. Proc., vol. 24, no. 12, pp. 2251–2262, 2016

work page 2016
[4]

Speech Enhancement Using a Minimum Mean-Square Error Short-Time Spectral Amplitude Estimator,

Y . Ephraim and D. Malah, “Speech Enhancement Using a Minimum Mean-Square Error Short-Time Spectral Amplitude Estimator,” IEEE Trans. Acoust., Speech, Signal Process., vol. ASSP-32, no. 6, pp. 1109– 1121, 1984

work page 1984
[5]

A novel fast nonstationary noise tracking approach based on mmse spectral power estimator,

Q. Zhang, M. Wang, Y . Lu, L. Zhang, and M. Idrees, “A novel fast nonstationary noise tracking approach based on mmse spectral power estimator,” Digital Signal Processing , vol. 88, pp. 41–52, 2019

2019
[6]

Fast nonstationary noise tracking based on log-spectral power mmse estimator and temporal recursive averaging,

Q. Zhang, M. Wang, Y . Lu, M. Idrees, and L. Zhang, “Fast nonstationary noise tracking based on log-spectral power mmse estimator and temporal recursive averaging,” IEEE Access, vol. 7, pp. 80 985–80 999, 2019

work page 2019
[7]

Real time speech enhancement in the waveform domain,

A. Defossez, G. Synnaeve, and Y . Adi, “Real time speech enhancement in the waveform domain,” in Proc. INTERSPEECH, 2020

2020
[8]

On loss functions for supervised monaural time-domain speech enhancement,

M. Kolbæk, Z.-H. Tan, S. H. Jensen, and J. Jensen, “On loss functions for supervised monaural time-domain speech enhancement,” IEEE/ACM Trans. Audio, speech, Lang. Process. , vol. 28, pp. 825–838, 2020

2020

Show all 34 references

[9]

DeepMMSE: A deep learning approach to mmse-based noise power spectral density estimation,

Q. Zhang, A. Nicolson, M. Wang, K. K. Paliwal, and C. Wang, “DeepMMSE: A deep learning approach to mmse-based noise power spectral density estimation,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 28, pp. 1404–1415, 2020

2020
[10]

Exploring tradeoffs in models for low- latency speech enhancement,

K. Wilson, M. Chinen, J. Thorpe, B. Patton, J. Hershey, R. A. Saurous, J. Skoglund, and R. F. Lyon, “Exploring tradeoffs in models for low- latency speech enhancement,” in Proc. IWAENC, 2018, pp. 366–370

2018
[11]

A regression approach to speech enhancement based on deep neural networks,

Y . Xu, J. Du, L.-R. Dai, and C.-H. Lee, “A regression approach to speech enhancement based on deep neural networks,” IEEE/ACM Trans. Audio, Speech, and Lang. Proc. , vol. 23, no. 1, pp. 7–19, 2014

2014
[12]

U-net: Convolutional networks for biomedical image segmentation,

O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical Image Computing and Computer Assisted Intervention , vol. 9351, 2015, pp. 234–241

2015
[13]

Dccrn: Deep complex convolution recurrent network for phase-aware speech enhancement,

Y . Hu, Y . Liu, and Lv, “Dccrn: Deep complex convolution recurrent network for phase-aware speech enhancement,” Proc. INTERSPEECH, pp. 2472–2476, 2020

2020
[14]

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 ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023, pp. 1–5

2023
[15]

Mamba in speech: Towards an alternative to self-attention,

X. Zhang, Q. Zhang, H. Liu, T. Xiao, X. Qian, B. Ahmed, E. Ambikaira- jah, H. Li, and J. Epps, “Mamba in speech: Towards an alternative to self-attention,” arXiv preprint arXiv:2405.12609 , 2024

2024 arXiv
[16]

Mp-senet: A speech enhancement model with parallel denoising of magnitude and phase spectra,

Y . Lu, Y . Ai, and Z. Ling, “Mp-senet: A speech enhancement model with parallel denoising of magnitude and phase spectra,” in Proc. Interspeech, N. Harte, J. Carson-Berndsen, and G. Jones, Eds., 2023, pp. 3834–3838

2023
[17]

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 The Thirty-Fourth AAAI Conference on Artificial Intelligence, 2020 . AAAI Press, 2020, pp. 9458–9465

2020
[18]

Complex ratio masking for monaural speech separation,

D. S. Williamson, Y . Wang, and D. Wang, “Complex ratio masking for monaural speech separation,” IEEE/ACM Trans. Audio, Speech, and Lang. Proc., vol. 24, no. 3, pp. 483–492, 2015

2015
[19]

Graph signal processing: overview, challenges, and appli- cations,

A. Ortega, P. Frossard, J. Kovacevic, J. M. F. Moura, and P. Van- dergheynst, “Graph signal processing: overview, challenges, and appli- cations,” Proc. IEEE, vol. 106, no. 5, pp. 808–828, 2018

2018
[20]

Speech signal processing on graphs: the graph frequency analysis and an improved graph wiener filtering method,

T. Wang, H. Guo, X. Yan, and Z. Yang, “Speech signal processing on graphs: the graph frequency analysis and an improved graph wiener filtering method,” Speech Commun., vol. 127, pp. 82–91, 2021

2021
[21]

Weighted speech distortion losses for neural-network-based real-time speech enhancement,

Y . Xia, S. Braun, C. K. A. Reddy, H. Dubey, R. Cutler, and I. Tashev, “Weighted speech distortion losses for neural-network-based real-time speech enhancement,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process., 2020, pp. 871–875

2020
[22]

DPCRN: dual-path convolution recurrent network for single channel speech enhancement,

X. Le, H. Chen, K. Chen, and J. Lu, “DPCRN: dual-path convolution recurrent network for single channel speech enhancement,” CoRR, vol. abs/2107.05429, 2021. [Online]. Available: https://arxiv.org/abs/2107. 05429

2021 arXiv
[23]

Dual application of speech enhancement for automatic speech recognition,

A. Pandey, C. Liu, Y . Wang, and Y . Saraf, “Dual application of speech enhancement for automatic speech recognition,” in IEEE Spoken Language Technology Workshop, 2021, pp. 223–228

2021
[24]

Multi-scale temporal frequency convolutional network with axial attention for multi-channel speech enhancement,

G. Zhang, C. Wang, L. Yu, and J. Wei, “Multi-scale temporal frequency convolutional network with axial attention for multi-channel speech enhancement,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process., 2022, pp. 9206–9210

2022
[25]

High fidelity speech enhancement with band-split rnn,

Y . L. Jianwei Yu, Hangting Chen, “High fidelity speech enhancement with band-split rnn,” 2024. [Online]. Available: https://arxiv.org/pdf/ 2212.00406

2024 arXiv
[26]

Single-channel speech enhancement using graph fourier transform,

C. Zhang and X. Pan, “Single-channel speech enhancement using graph fourier transform,” in Proc. INTERSPEECH, 2022, pp. 946–950

2022
[27]

Time-domain speech sep- aration networks with graph encoding auxiliary,

T. Wang, Z. Pan, M. Ge, Z. Yang, and H. Li, “Time-domain speech sep- aration networks with graph encoding auxiliary,” IEEE Signal Process. Lett., vol. 30, pp. 110–114, 2023

2023
[28]

DCCRN+: channel-wise subband DCCRN with SNR estimation for speech enhancement,

S. Lv, Y . Hu, S. Zhang, and L. Xie, “DCCRN+: channel-wise subband DCCRN with SNR estimation for speech enhancement,” in Proc. IN- TERSPEECH, 2021, pp. 2816–2820

2021
[29]

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

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

2020
[30]

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. Int. Conf. Oriental COCOSDA, Nov 2013 , 2013, pp. 1–4

2013
[31]

The diverse environments multi- channel acoustic noise database: A database of multichannel environ- mental noise recordings,

J. Thiemann, N. Ito, and E. Vincent, “The diverse environments multi- channel acoustic noise database: A database of multichannel environ- mental noise recordings,” in Acoust. Soc. Am. , vol. 133, no. 5, 2013, p. 3591–3591

2013
[32]

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. Hekstra, “Perceptual evaluation of speech quality (pesq)-a new method for speech quality assessment of telephone networks and codecs,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process. , 2001, pp. 749–752

2001
[33]

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. IEEE Int. Conf. Acoust., Speech, Signal Process., 2010, pp. 4214–4217

2010
[34]

SDR - half- baked or well done?

J. L. Roux, S. Wisdom, H. Erdogan, and J. R. Hershey, “SDR - half- baked or well done?” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process., 2019, pp. 626–630

2019

Pith tools

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

[1] [1]

P. C. Loizou, Speech Enhancement: Theory and Practice, 2nd ed. Boca Raton, FL, USA: CRC Press, Inc., 2013

work page 2013

[2] [2]

Supervised speech separation based on deep learning: An overview,

D. Wang and J. Chen, “Supervised speech separation based on deep learning: An overview,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 26, no. 10, pp. 1702–1726, 2018

2018

[3] [3]

On MMSE-based estimation of amplitude and complex speech spectral coefficients under phase- uncertainty,

M. Krawczyk-Becker and T. Gerkmann, “On MMSE-based estimation of amplitude and complex speech spectral coefficients under phase- uncertainty,” IEEE/ACM Trans. Audio, Speech, and Lang. Proc., vol. 24, no. 12, pp. 2251–2262, 2016

work page 2016

[4] [4]

Speech Enhancement Using a Minimum Mean-Square Error Short-Time Spectral Amplitude Estimator,

Y . Ephraim and D. Malah, “Speech Enhancement Using a Minimum Mean-Square Error Short-Time Spectral Amplitude Estimator,” IEEE Trans. Acoust., Speech, Signal Process., vol. ASSP-32, no. 6, pp. 1109– 1121, 1984

work page 1984

[5] [5]

A novel fast nonstationary noise tracking approach based on mmse spectral power estimator,

Q. Zhang, M. Wang, Y . Lu, L. Zhang, and M. Idrees, “A novel fast nonstationary noise tracking approach based on mmse spectral power estimator,” Digital Signal Processing , vol. 88, pp. 41–52, 2019

2019

[6] [6]

Fast nonstationary noise tracking based on log-spectral power mmse estimator and temporal recursive averaging,

Q. Zhang, M. Wang, Y . Lu, M. Idrees, and L. Zhang, “Fast nonstationary noise tracking based on log-spectral power mmse estimator and temporal recursive averaging,” IEEE Access, vol. 7, pp. 80 985–80 999, 2019

work page 2019

[7] [7]

Real time speech enhancement in the waveform domain,

A. Defossez, G. Synnaeve, and Y . Adi, “Real time speech enhancement in the waveform domain,” in Proc. INTERSPEECH, 2020

2020

[8] [8]

On loss functions for supervised monaural time-domain speech enhancement,

M. Kolbæk, Z.-H. Tan, S. H. Jensen, and J. Jensen, “On loss functions for supervised monaural time-domain speech enhancement,” IEEE/ACM Trans. Audio, speech, Lang. Process. , vol. 28, pp. 825–838, 2020

2020