Pith. sign in

REVIEW 4 major objections 6 minor 54 references

BSDB-Net: Band-Split Dual-Branch Network with Selective State Spaces Mechanism for Monaural 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 This paper claims a dual-branch, band-split network using Mamba sequence modeling can cut computational cost of monaural speech enhancement by an average factor of 8.3 while matching or exceeding state-of-the-art quality.

desk verdict A genuinely cheap causal SE architecture worth knowing about, but the abstract's complexity/perfomance claim is overstated and needs fixing before this is citable. read the letter →

arxiv 2412.19099 v1 pith:YTEMYJGW submitted 2024-12-26 cs.SD eess.AS

Cunhang Fan , Enrui Liu , Andong Li , Jianhua Tao , Jian Zhou , Jiahao Li , Chengshi Zheng , Zhao Lv This is my paper · ORCID
classification cs.SDeess.AS
keywords speechenhancementmonauralMambaselectivestatespacemodelsband-splitdual-branchamplitude-phasedecouplingcomplexspectralmapping
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 aims to solve two problems at once in monaural speech enhancement: the compensation effect caused by jointly estimating amplitude and phase from complex spectra, and the high computational cost of modern enhancement models. It proposes a dual-branch network in which one branch enhances the magnitude spectrum while the other estimates the complex spectrum, with an interaction module that lets each branch suppress or supply what the other lacks. To cut cost, it groups frequency bins into non-overlapping bands and models the resulting sequences with Mamba, a selective state-space model that runs in linear time. On WSJ0-SI84 and VoiceBank+Demand, the authors report matching or better quality than earlier methods while cutting average complexity by 8.3 times relative to baselines and 25 times relative to transformer-based models. The payoff, if right, is a path to high-quality speech enhancement that fits on devices with tight compute budgets.

What carries the argument

Three components carry the argument. The Band-Split module compresses the F frequency bins into K non-overlapping bands, each projected to dimension N, producing a compact tensor for sequence modeling, and the Mask-Decoder merges bands back into a full-spectrum mask. The dual-branch structure consists of a Magnitude Enhancement Network and a Complex Spectral Enhancement Network running in parallel, with an Interaction module that concatenates the two streams, creates a sigmoid mask, and lets one branch repair missing components of the other. The Mamba-Block performs bidirectional frequency modeling and unidirectional causal time modeling using selective state-space equations, giving linear complexity in both axes.

What would settle it

Replace the band-split in BSDB-Net with the full FFT resolution (K equal to the number of frequency bins) on WSJ0-SI84 Set-B while keeping everything else fixed; if PESQ, ESTOI, or SI-SDR improve by a large margin, the compression is discarding information the enhancement depends on. Independently recomputing the Table 2 MACs on one-second inputs with standard profiling would confirm the 8.3x and 25x reduction ratios.

Watch

Extended reading notes

Core claim

The central claim is that decoupling magnitude and phase in parallel branches, rather than mapping a single complex spectrum to another one, removes the compensation effect in which amplitude is degraded to repair phase. The magnitude branch applies a mask to roughly suppress noise; the complex branch estimates real and imaginary components; and the two streams are combined after an interaction step. Band-splitting compresses the frequency dimension before sequence modeling, and Mamba-based blocks model time and frequency with linear complexity while keeping the model causal. With its chosen configuration, the paper reports quality scores above all compared baselines on WSJ0-SI84 and VoiceBank+Demand while running at 1.68 G/s MACs, and states that this is state-of-the-art across all metrics.

Load-bearing premise

The load-bearing premise is that grouping frequency bins into K fixed non-overlapping bands loses no information the enhancement needs, since the paper tests band-splitting but not a no-band-split version.

Editorial extensions

If this is right

  • Speech enhancement can be pushed toward real-time, low-power deployment because the model operates at megabyte-scale per second rather than gigabyte-scale per second.
  • Causal, streaming enhancement is compatible with this design because the time-dimension Mamba is unidirectional.
  • Scaling behavior favors depth over width, so a smaller hidden dimension with more stacked Mamba blocks can be chosen to balance quality and cost.
  • Replacing quadratic-attention transformers with Mamba in other time-frequency sequence tasks may produce large complexity savings with only a small quality difference.

Reading between the lines

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

  • The published 8.3x average is computed over the chosen baselines and excludes the non-causal transformer model DBTNet; extending the comparison to that model would likely lower the headline ratio.
  • The band-split count K is a tunable knob the paper does not explore in depth; finer bands would probably help difficult, phase-heavy conditions at higher cost.
  • The interaction module's specific contribution is not isolated in the ablations, so its role in suppressing noise versus restoring signal remains open.
  • The dual-branch decoupling idea may transfer directly to other speech tasks such as dereverberation and target extraction, which the paper lists as future work.
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 / 6 minor

Summary. The paper proposes BSDB-Net, a causal monaural speech enhancement model that decouples magnitude and complex-spectrum enhancement in two parallel branches, compresses the frequency axis with a band-split module, and uses Mamba-based blocks for time and frequency sequence modeling. The method is evaluated on a WSJ0-SI84+DNS-Challenge corpus and on VoiceBank+Demand. The authors report state-of-the-art or competitive PESQ, ESTOI, and SI-SDR results and claim an average 8.3x reduction in computational complexity relative to baselines, with the best 128-6 configuration at 1.68 G/s and PESQ 2.92.

Significance. If the reported numbers are taken at face value after correcting the Table 1 typo, the paper makes a useful contribution to efficient speech enhancement: it demonstrates that a causal dual-branch architecture with band-split frequency compression and Mamba sequence modeling can reach the performance level of much heavier T-F-domain models on WSJ0-SI84 and VoiceBank+Demand. The ablations of the dual-branch structure and the Mamba-versus-LSTM/Transformer replacement are informative, and the complexity table using a fixed one-second input is a useful step toward fair MAC comparisons. The central efficiency claim, however, is currently overstated relative to the evidence in Tables 1 and 2, so the significance depends on how the authors choose to scope the claim.

major comments (4)
  1. [Table 1, Set-A, BSDB row] The BSDB(ours) PESQ values for Set-A are 2.43, 2.92, and 2.26 at -5, 0, and 5 dB, with an average reported as 2.87; arithmetic requires the 5 dB value to be approximately 3.26, not 2.26. Because the row average and the subsequent state-of-the-art claim depend on this value, please correct the cell and re-verify all metrics in the table.
  2. [Abstract and Table 2] The claim of an average 8.3x complexity reduction while maintaining superior performance is not supported when the comparison is restricted to baselines with comparable PESQ. The strongest baseline in Table 2, DBTNet, reaches PESQ 3.18 at 42.64 G/s whereas BSDB-Net (128-6) reaches PESQ 2.92 at 1.68 G/s, a 0.26 PESQ loss; against GaGNet (2.85 PESQ, 2.81 G/s) and CTSNet (2.79 PESQ, 5.48 G/s) the reduction is roughly 1.7-3.3x, not 8.3x. Please report the complexity ratio relative to comparable-performance baselines and align the abstract wording with the actual comparison set, since Table 1 omits DBTNet entirely.
  3. [Ablation Study and Eqs. (2)-(4)] The ablation section covers the dual-branch structure, the number of Mamba layers and hidden dimensions, and Mamba versus LSTM/Transformer, but it does not test the band-split compression itself. Because band-splitting is one of the two mechanisms behind the claimed complexity reduction, an experiment that varies K or bypasses band-splitting is needed to show that the frequency compression does not discard information necessary for phase recovery; without it, the performance-complexity trade-off cannot be attributed to the proposed design.
  4. [The Number of Layers and Hidden Layers of Mamba-Block and Table 2] The text states that the model compresses computational complexity 'to the level of Mb/s', but Table 2 reports all BSDB-Net variants between 0.88 G/s and 4.26 G/s. Please correct the units, and also make the complexity claim numerically consistent with the abstract's '8.3 times' by reporting the exact average and the baseline set over which it is computed.
minor comments (6)
  1. [Table 1, Set-B, noisy row] The noisy PESQ value at 0 dB is written as '2,04' and should be '2.04'.
  2. [Table 2] The heading 'Modle' should be 'Model', and the DBTNet row uses '%' in the Cau. column instead of a check mark; the text should explicitly state that DBTNet is non-causal.
  3. [Proposed Architecture] The text refers to the 'BSDS framework' where 'BSDB' is intended.
  4. [References] The reference 'Chan, J. N. L. Q., W. 2016' is malformed and should give the full author list.
  5. [Eqs. (17)-(18)] The SSM recurrence is stated without the discretization equations; please include them or cite the Mamba paper at that point.
  6. [Tables 1 and 4] Please report standard deviations or significance tests for the main metrics, since several reported differences are small and the state-of-the-art claim rests on them; also standardize the names 'CTS-Net'/'CSTnet' and 'S4NDUNet'/'S4DSE'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: BSDB-Net's results are benchmark evaluations of a new architecture, not derived from their own inputs.

full rationale

The paper's derivation chain is architectural and empirical, not definitional. Equations (1)-(29) define a mapping from noisy STFT spectra through band-splitting, dual-branch Mamba blocks, interaction, and mask decoding; no target metric is used as an input to any equation. The band-split in Eqs. (2)-(4) is a preprocessing choice adopted from Yu and Luo (2023), and the Mamba formulation in Eqs. (17)-(18) is the standard selective SSM; neither is defined in terms of the reported PESQ/ESTOI/SI-SDR results. Complexity figures (Table 2) are measured MACs on fixed-length inputs, an independent calculation, and the performance numbers come from standard public benchmarks (WSJ0-SI84, VoiceBank+Demand). Self-citations (e.g., CompNet, CTSNet, GaGNet) appear only as related-work baselines or architecture inspiration, not as load-bearing justification for the paper's central claims. The claim that the 8.3x complexity reduction is averaged over a set including heavier baselines and that the model trails DBTNet by 0.26 PESQ is a legitimate concern about how 'superior performance' is framed, but it concerns comparison-set selection and claim support, not circularity: no quantity in the paper reduces to its own output by construction. The absence of a no-band-split ablation is an empirical coverage gap, not a circular step.

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

This is an empirical architecture paper with no mathematical derivation. The central claim rests on standard SE assumptions (additive STFT noise, metric validity) and on design hyperparameters selected via ablation. The custom dataset is not released, which limits reproducibility.

free parameters (4)
  • Mamba input hidden dimension = 128 (selected via ablation; 64 and 256 also evaluated)
    The dimension of the input data to Mamba affects performance and complexity; the paper sweeps 64, 128, 256 and selects 128-6 as the balanced configuration (Table 2).
  • Number of stacked Mamba blocks = 6 (in the 128-6 configuration; 2 and 4 also evaluated)
    Stacking depth is ablated in Table 2; deeper stacks improve PESQ at higher MACs.
  • Loss weighting beta = 0.5
    Set empirically in Eq. (29): L = beta*L_RI + (1-beta)*L_Mag, with beta=0.5.
  • Number of frequency bands K = not specified in text
    The band-split module uses K bands but K is not stated; this is a free design choice affecting frequency compression.
assumptions (4)
  • domain assumption Additive noise model: Y(t,f) = X(t,f) + N(t,f) (Eq. 1)
    The paper models noisy speech as clean plus noise in the STFT domain, the standard SE assumption.
  • domain assumption Power compression |X|^0.5 and |S|^0.5 improves learnability
    Follows Li et al. (2021b) with compression coefficient 0.5; the paper does not re-justify it.
  • standard math State-space discretization (Eqs. 17-18) is valid for speech sequence modeling
    The Mamba discretization is taken from Gu and Dao (2023) without re-derivation.
  • domain assumption Objective metrics (PESQ, ESTOI, SI-SDR, MOS) reflect perceived quality
    Evaluation uses standard metrics from the SE literature; no listening tests are reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of BSDB-Net: Band-Split Dual-Branch Network with Selective State Spaces Mechanism for Monaural Speech Enhancement." pith.science (2026). https://pith.science/paper/YTEMYJGW

@misc{pith2026241219099,
  author       = {Pith},
  title        = {Pith review of: BSDB-Net: Band-Split Dual-Branch Network with Selective State Spaces Mechanism for Monaural Speech Enhancement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YTEMYJGW}},
  note         = {Machine review of arXiv:2412.19099}
}
read the original abstract

Although the complex spectrum-based speech enhancement(SE) methods have achieved significant performance, coupling amplitude and phase can lead to a compensation effect, where amplitude information is sacrificed to compensate for the phase that is harmful to SE. In addition, to further improve the performance of SE, many modules are stacked onto SE, resulting in increased model complexity that limits the application of SE. To address these problems, we proposed a dual-path network based on compressed frequency using Mamba. First, we extract amplitude and phase information through parallel dual branches. This approach leverages structured complex spectra to implicitly capture phase information and solves the compensation effect by decoupling amplitude and phase, and the network incorporates an interaction module to suppress unnecessary parts and recover missing components from the other branch. Second, to reduce network complexity, the network introduces a band-split strategy to compress the frequency dimension. To further reduce complexity while maintaining good performance, we designed a Mamba-based module that models the time and frequency dimensions under linear complexity. Finally, compared to baselines, our model achieves an average 8.3 times reduction in computational complexity while maintaining superior performance. Furthermore, it achieves a 25 times reduction in complexity compared to transformer-based models.

Figures

Figures reproduced from arXiv: 2412.19099 by the authors.

Figure 1
Figure 1. Overall architecture of the proposed BSDB-Net consists of three main components. The first part includes the Band [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. (a) The Band-Split module divides frequency [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The Mamba-Block: It is primarily divided into [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 45 canonical work pages

  1. [1]

    Alex, T.; Ahmed, S.; Mustafa, A.; Awais, M.; and Jackson, P. J. 2024. DTF-AT: Decoupled Time-Frequency Audio Transformer for Event Classification. In AAAI 2024, volume 38, 17647--17655

    work page 2024

  2. [2]

    Z.; and Koltun, V

    Bai, S.; Kolter, J. Z.; and Koltun, V. 2018. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. arXiv preprint arXiv:1803.01271

    arXiv 2018

  3. [3]

    V.; Wang, X.; Takaki, S.; and Yamagishi, J

    Botinhao, C. V.; Wang, X.; Takaki, S.; and Yamagishi, J. 2016. Investigating RNN-based speech enhancement methods for noise-robust Text-to-Speech. In 9th ISCA Speech Synthesis Workshop, 159--165

    work page 2016

  4. [4]

    Chan, J. N. L. Q., W. 2016. Listen, attend and spell: A neural network for large vocabulary conversational speech recognition . In speech and signal processing, 4960–4964. Detroit, Michigan: IEEE international conference on acoustics

    work page 2016

  5. [5]

    E.; Wang, D.; and Healy, E

    Chen, J.; Wang, Y.; Yoho, S. E.; Wang, D.; and Healy, E. W. 2016. Large-scale training to increase speech intelligibility for hearing-impaired listeners in novel noises. The Journal of the Acoustical Society of America, 139(5): 2604--2612

    work page 2016

  6. [6]

    Divenyi, P. 2004. Speech separation by humans and machines. Springer Science & Business Media

    work page 2004

  7. [7]

    Fan, C.; Zhang, H.; Li, A.; Xiang, W.; Zheng, C.; Lv, Z.; and Wu, X. 2023. CompNet: Complementary network for single-channel speech enhancement. Neural Networks, 168: 508--517

    work page 2023

  8. [8]

    Fu, S.-W.; Liao, C.-F.; Tsao, Y.; and Lin, S.-D. 2019. Metricgan: Generative adversarial networks based black-box metric scores optimization for speech enhancement. In International Conference on Machine Learning, 2031--2041. PmLR

    work page 2019

Show all 54 references
  1. [9]

    Gu, A.; and Dao, T. 2023. Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752

    2023 arXiv

  2. [10]

    Gu, A.; Dao, T.; Ermon, S.; Rudra, A.; and R \'e , C. 2020. Hippo: Recurrent memory with optimal polynomial projections. Advances in neural information processing systems, 33: 1474--1487

    2020

  3. [11]

    Gu, A.; Goel, K.; and R \'e , C. 2021. Efficiently modeling long sequences with structured state spaces. International Conference on Learning Representations

    2021

  4. [12]

    Hamilton, J. D. 1994. State-space models. Handbook of econometrics, 4: 3039--3080

    1994

  5. [13]

    Hao, X.; Su, X.; Horaud, R.; and Li, X. 2021. Fullsubnet: A full-band and sub-band fusion model for real-time single-channel speech enhancement. In ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing, 6633--6637. IEEE

    2021

  6. [14]

    Hu, Y.; Liu, Y.; Lv, S.; Xing, M.; Zhang, S.; Fu, Y.; Wu, J.; Zhang, B.; and Xie, L. 2020. DCCRN: Deep complex convolution recurrent network for phase-aware speech enhancement. In INTERSPEECH, volume 2020. International Speech Communication Association

    2020

  7. [15]

    Hu, Y.; and Loizou, P. C. 2007. Evaluation of objective quality measures for speech enhancement. IEEE Transactions on audio, speech, and language processing, 16(1): 229--238

    2007

  8. [16]

    Hummersone, C.; Stokes, T.; and Brookes, T. 2014. On the ideal ratio mask as the goal of computational auditory scene analysis. In Blind source separation: advances in theory, algorithms and applications, 349--368. Springer

    2014

  9. [17]

    Jensen, J.; and Taal, C. H. 2016. An algorithm for predicting the intelligibility of speech masked by modulated noise maskers. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 24(11): 2009--2022

    2016

  10. [18]

    Koizumi, Y.; Yatabe, K.; Delcroix, M.; Masuyama, Y.; and Takeuchi, D. 2020. Speech enhancement using self-adaptation and multi-head self-attention. In ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing, 181--185. IEEE

    2020

  11. [19]

    Li, A.; Liu, W.; Zheng, C.; Fan, C.; and Li, X. 2021 a . Two heads are better than one: A two-stage complex spectral mapping approach for monaural speech enhancement. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 29: 1829--1843

    2021

  12. [20]

    Li, A.; You, S.; Yu, G.; Zheng, C.; and Li, X. 2022 a . Taylor, Can You Hear Me Now? A Taylor-Unfolding Framework for Monaural Speech Enhancement. In Raedt, L. D., ed., Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, IJCAI-22 , 4193--...

    2022

  13. [21]

    Li, A.; Zheng, C.; Peng, R.; and Li, X. 2021 b . On the importance of power compression and phase estimation in monaural speech dereverberation. JASA express letters, 1(1)

    2021

  14. [22]

    Li, A.; Zheng, C.; Zhang, L.; and Li, X. 2022 b . Glance and gaze: A collaborative learning framework for single-channel speech enhancement. Applied Acoustics, 187: 108499

    2022

  15. [23]

    Li, K.; and Chen, G. 2024. Spmamba: State-space model is all you need in speech separation. arXiv preprint arXiv:2404.02063

    2024 arXiv

  16. [24]

    Liu, G.; Shi, J.; Chen, X.; Xu, J.; and Xu, B. 2018. Improving Speech Separation with Adversarial Network and Reinforcement Learning. In 2018 IJCNN, 1--7

    2018

  17. [25]

    Lu, Y.-X.; Ai, Y.; and Ling, Z.-H. 2023. MP-SENet: A Speech Enhancement Model with Parallel Denoising of Magnitude and Phase Spectra . In Proc. INTERSPEECH 2023, 3834--3838

    2023

  18. [26]

    Luo, Y.; Chen, Z.; and Yoshioka, T. 2020. Dual-path rnn: efficient long sequence modeling for time-domain single-channel speech separation. In ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing, 46--50. IEEE

    2020

  19. [27]

    Luo, Y.; and Mesgarani, N. 2019. Conv-tasnet: Surpassing ideal time--frequency magnitude masking for speech separation. IEEE/ACM transactions on audio, speech, and language processing, 27(8): 1256--1266

    2019

  20. [28]

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

    2016 arXiv

  21. [29]

    P-J, K.; Yang, C.; Siniscalchi, S.; et al. 2023. A Multi-dimensional Deep Structured State Space Approach to Speech Enhancement Using Small-footprint Models. In INTERSPEECH, volume 2023, 2453--2457. International Speech Communication Association

    2023

  22. [30]

    Pandey, A.; and Wang, D. 2019. TCNN: Temporal convolutional neural network for real-time speech enhancement in the time domain. In ICASSP 2019, 6875--6879. IEEE

    2019

  23. [31]

    Pandey, A.; and Wang, D. 2020. Densely connected neural network with dilated convolutions for real-time speech enhancement in the time domain. In ICASSP 2020, 6629--6633. IEEE

    2020

  24. [32]

    Pascual, S.; Bonafonte, A.; and Serra, J. 2017. SEGAN: Speech enhancement generative adversarial network. arXiv preprint arXiv:1703.09452

    2017 arXiv

  25. [33]

    B.; and Baker, J

    Paul, D. B.; and Baker, J. 1992. The design for the Wall Street Journal-based CSR corpus. In Speech and Natural Language: Proceedings of a Workshop Held at Harriman, New York, February 23-26, 1992

    1992

  26. [34]

    Qiu, Z.; Fu, M.; Yu, Y.; Yin, L.; Sun, F.; and Huang, H. 2023. SRTNET: Time Domain Speech Enhancement via Stochastic Refinement. In ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 1--5

    2023

  27. [35]

    K.; Gopal, V.; Cutler, R.; Beyrami, E.; Cheng, R.; Dubey, H.; Matusevych, S.; Aichner, R.; Aazami, A.; Braun, S.; et al

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

    2020

  28. [36]

    W.; Beerends, J

    Rix, A. W.; Beerends, J. G.; Hollier, M. P.; and Hekstra, A. P. 2001. Perceptual evaluation of speech quality (PESQ)-a new method for speech quality assessment of telephone networks and codecs. In 2001 IEEE international conference on acoustics, speech, and signal processing. ...

    2001

  29. [37]

    H.; Shah, N.; and Patil, H

    Soni, M. H.; Shah, N.; and Patil, H. A. 2018. Time-frequency masking-based speech enhancement using generative adversarial network. In ICASSP 2018 - 2018 IEEE International Conference on Acoustics, Speech and Signal Processing, 5039--5043. IEEE

    2018

  30. [38]

    Tan, K.; and Wang, D. 2018. A convolutional recurrent neural network for real-time speech enhancement. In Interspeech, volume 2018, 3229--3233

    2018

  31. [39]

    Tan, K.; and Wang, D. 2019. Learning complex spectral mapping with gated convolutional recurrent networks for monaural speech enhancement. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 28: 380--390

    2019

  32. [40]

    Thiemann, J.; Ito, N.; and Vincent, E. 2013. The diverse environments multi-channel acoustic noise database (demand): A database of multichannel environmental noise recordings. In Proceedings of Meetings on Acoustics, volume 19. AIP Publishing

    2013

  33. [41]

    Veaux, C.; Yamagishi, J.; and King, S. 2013. The voice bank corpus: Design, collection and data analysis of a large regional accent speech database. In 2013 international conference oriental COCOSDA held jointly with 2013 conference on Asian spoken language research and evalua...

    2013

  34. [42]

    Wang, K.; He, B.; and Zhu, W.-P. 2021. TSTNN: Two-stage transformer based neural network for speech enhancement in the time domain. In ICASSP 2021 IEEE International Conference on Acoustics, Speech and Signal Processing, 7098--7102. IEEE

    2021

  35. [43]

    Wang, Z.-Q.; Wichern, G.; and Le Roux, J. 2021. On the compensation between magnitude and phase in speech separation. IEEE Signal Processing Letters, 28: 2018--2022

    2021

  36. [44]

    L.; and Meyer, B

    Westhausen, N. L.; and Meyer, B. T. 2020. Dual-Signal Transformation LSTM Network for Real-Time Noise Suppression. Interspeech 2020

    2020

  37. [45]

    S.; Wang, Y.; and Wang, D

    Williamson, D. S.; Wang, Y.; and Wang, D. 2015. Complex ratio masking for monaural speech separation. IEEE/ACM transactions on audio, speech, and language processing, 24(3): 483--492

    2015

  38. [46]

    Xu, X.; Tu, W.; and Yang, Y. 2023. Selector-enhancer: learning dynamic selection of local and non-local attention operation for speech enhancement. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, 13853--13860

    2023

  39. [47]

    Yin, D.; Luo, C.; Xiong, Z.; and Zeng, W. 2020. Phasen: A phase-and-harmonics-aware speech enhancement network. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, 9458--9465

    2020

  40. [48]

    Yu, G.; Li, A.; Wang, H.; Wang, Y.; Ke, Y.; and Zheng, C. 2022. DBT-Net: Dual-branch federative magnitude and phase estimation with attention-in-attention transformer for monaural speech enhancement. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 30: 2629--2644

    2022

  41. [49]

    Yu, J.; and Luo, Y. 2023. Efficient Monaural Speech Enhancement with Universal Sample Rate Band-Split RNN. In ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 1--5

    2023

  42. [50]

    Yue, H.; Duo, W.; Peng, X.; and Yang, J. 2022. Reference-based speech enhancement via feature alignment and fusion network. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, 11648--11656

    2022

  43. [51]

    Zaremba, W.; Sutskever, I.; and Vinyals, O. 2014. Recurrent neural network regularization. arXiv preprint arXiv:1409.2329

    2014 arXiv

  44. [52]

    Zhang, Y.; Zou, H.; and Zhu, J. 2024. A Two-Stage Framework in Cross-Spectrum Domain for Real-Time Speech Enhancement. In ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 12587--12591. IEEE

    2024

  45. [53]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  46. [54]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

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