Pith. sign in

REVIEW 3 major objections 7 minor 31 references

U-Mamba-Net: A highly efficient Mamba-based U-net style network for noisy and reverberant speech separation

T0 review · 3 major / 7 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 The paper claims that a lightweight U-Mamba-Net separates two speakers from noisy, reverberant mixtures with higher SNR-family scores than prior one-task and cascade models while using roughly one-sixteenth the compute.

desk verdict The efficiency story is solid; the accuracy comparison depends on an unstated baseline retraining protocol, and the paper's own perceptual numbers undercut its headline claim. read the letter →

arxiv 2412.18217 v1 pith:WIH3ZOJT submitted 2024-12-24 cs.SD cs.LGeess.AS

Shaoxiang Dang , Tetsuya Matsumoto , Yoshinori Takeuchi , Hiroaki Kudo This is my paper · ORCID
classification cs.SDcs.LGeess.AS
keywords speechseparationnoisyandreverberantMambastatespacemodelsU-NetefficientneuralnetworksLibri2mixsingle-channelsource
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes U-Mamba-Net, a compact single-task model for separating two speakers from noisy and reverberant mixtures, and reports that it outperforms earlier one-task models and a cascaded two-stage baseline on SNR-family metrics while using far less computation. The key architectural move is to alternate a U-Net, which learns multi-resolution local features cheaply, with a Mamba module, a selective state space model that carries long-range dependencies in linear time. On a simulated noisy-reverberant Libri2mix test set, U-Mamba-Net reaches an SI-SNR improvement of 8.50 dB with 4.4 million parameters and 2.5 GMACs, against 8.08 dB and 40.2 GMACs for DPRNN-CMTL. The authors do not claim it wins every metric: a cascade DPRNN still leads on perceptual and denoising scores. A sympathetic reading of the contribution is an efficiency-and-quality trade-off, with separation metrics near the top of the table at a fraction of the compute.

What carries the argument

The load-bearing object is the U-Mamba block: a U-Net module, a compact fully convolutional contracting-and-expanding path with skip connections, followed by a Mamba module, a selective structured state space model that decides what to keep in its hidden state, runs in linear time, and is initialized with HiPPO orthogonal-polynomial projections. The U-Net supplies multi-resolution local feature filtering; Mamba supplies global, input-dependent selection over the long sequence; and a residual connection around Mamba stabilizes training. Repeating this alternation across sixteen stacked blocks is how the model captures long-term dependencies without quadratic self-attention.

What would settle it

Retrain DPRNN and DPRNN-CMTL on the exact same simulated noisy-reverberant training data, with the same three-second chunks and 8 kHz sampling, then compare SI-SNRi on the same test split; if either model reaches 8.50 dB or higher, the paper's central efficiency-advantage claim loses its quantitative support.

Watch

Extended reading notes

Core claim

The central claim is that a single end-to-end network can beat cascaded multi-task models on separation-quality metrics in noisy and reverberant conditions by making Mamba act as a feature filter that alternates with U-Net blocks. U-Mamba-Net uses a one-dimensional convolutional encoder to map the waveform to a time-frequency-like representation, passes it through sixteen stacked U-Mamba blocks, estimates masks, and decodes with a transposed convolution. Each block first runs a U-Net with four down-and-up sampling depths and residual connections, then runs Mamba on the result, then adds the Mamba output back as a residual. With this design the paper reports SI-SNRi of 8.50 dB, SDRi of 8.62 dB, and SIRi of 17.67 dB on its simulated noisy-reverberant Libri2mix, beating DPRNN and DPRNN-CMTL while using 2.5 GMACs versus 23.9 and 40.2.

Load-bearing premise

The reported advantage over DPRNN and Conv-TasNet assumes those baselines were measured on the same noisy-reverberant, 8 kHz Libri2mix setup, and the paper does not describe how the baseline numbers were produced.

Editorial extensions

If this is right

  • If the reported results hold, a single-task end-to-end separator can match or beat cascade pipelines on SNR-family metrics at much lower compute, weakening the assumption that noisy-reverberant separation requires an explicit denoising sub-task.
  • The architecture provides a template for cheap long-context separation: replace dual-path RNN or transformer blocks in existing time-domain separators with a U-Net-plus-Mamba alternation.
  • Because Mamba's cost is linear in sequence length, the model should scale to longer utterances without the quadratic blow-up of self-attention, though the paper only tests three-second inputs.
  • The ablation studies show feature dimension is the dominant performance lever, with F=192 raising SI-SNRi to 8.85 dB at 9.7 million parameters and 5.3 GMACs.
  • SNR improvement and perceptual quality do not move together, since DPRNN-CMTL still leads on PESQ, CSIG, CBAK, and COVL, so a practical deployment may need to choose which objective matters more.

Reading between the lines

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

  • The comparison in Table III does not state whether baselines were retrained on the same noisy-reverberant, 8 kHz protocol; if those numbers come from models trained on the standard anechoic Libri2mix, a fair retraining could erode or erase the 0.42 dB gap over DPRNN-CMTL.
  • The same alternating U-Net-plus-selective-SSM design could be dropped into other convolutional source separators, such as music separation or target-speaker extraction, where the local-feature and long-dependency split is similar.
  • The authors' own listening observation that harmonics are less clear suggests SNR metrics understate a perceptual weakness of the Mamba path; a subjective listening test or a distortion-focused metric would make the trade-off explicit.
  • Testing on a standard public noisy-reverberant benchmark rather than a self-simulated version would reveal whether the efficiency advantage survives out-of-distribution room and noise conditions.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. The paper proposes U-Mamba-Net, a lightweight time-domain speech separation model that alternates U-Net blocks with Mamba (selective state space) modules within an encoder-mask-decoder architecture. The model is trained with permutation-invariant SI-SNR loss and evaluated on a simulated noisy-reverberant version of Libri2mix at 8 kHz, with RIRs generated by Pyroomacoustics (T60 0.2-0.6 s) and WHAM! noise. The main results (Table III) report an SI-SNRi of 8.50 dB with 4.4 M parameters and 2.5 GMACs, outperforming DPRNN (7.59 dB) and a DPRNN-based cascade multi-task system (8.08 dB), while using substantially less compute. The paper also presents perceptual and denoising metrics (Table IV), ablations over feature dimension, block count, depth, and upsampling method, and a spectrogram visualization.

Significance. If the reported comparison is valid, the paper's efficiency claim is significant: a single-task end-to-end model achieving competitive or better SNR-oriented separation performance than a cascade multi-task baseline at roughly one-sixteenth the computational cost would be a useful contribution for the speech separation community. The paper is transparent about its weaknesses, noting that DPRNN-CMTL retains an advantage on several perceptual and denoising metrics. The Mamba formulation (Eqs. 1-9) is standard and correctly presented, the training protocol uses a public dataset with standard SI-SNR PIT, and the ablations cover the main architectural hyperparameters. However, the central comparison to previous models depends on baseline numbers whose evaluation protocol is not stated, which is a load-bearing omission that must be addressed before the results can be accepted.

major comments (3)
  1. [Section III.A, III.C, Table III] The manuscript does not state whether the baseline models (TasNet, SuDoRM-RF, Conv-TasNet, DPRNN, DPRNN-CMTL) in Table III were retrained and evaluated under the same noisy-reverberant 8 kHz simulation (T60 0.2-0.6 s, WHAM! noise) used for U-Mamba-Net. If these numbers are quoted from the original publications, which typically use anechoic Libri2mix at 16 kHz, then the reported SI-SNRi advantages of 0.42-0.92 dB over DPRNN and DPRNN-CMTL, and the comparisons in Table IV, confound method quality with evaluation conditions. The authors must specify exactly how each baseline was obtained, and if the baselines were not retrained under the identical protocol, the experiments must be redone with common data and preprocessing.
  2. [Section IV.A, Table III] The DPRNN-CMTL baseline is implemented with the authors' 'Enhancement Priority Pipeline' (reference [3]), but the paper does not describe the EPP training schedule, intermediate supervision labels, number of parameters attributable to the pipeline, or any differences from the original CMTL formulation. Because the claim of surpassing a cascade multi-task system on SI-SNRi rests directly on this baseline, the exact implementation details must be documented, and the authors should clarify whether the reported 5.6 M parameters and 40.2 GMACs include the entire pipeline.
  3. [Section IV.A, Tables III and IV] No statistical significance, error bars, or multiple-seed variance are reported for any result. The observed improvements of 0.42-0.92 dB over DPRNN and DPRNN-CMTL may be within run-to-run variability for a single training seed; the authors should provide means and standard deviations over at least three seeds for the main comparisons, or otherwise justify that the differences are significant.
minor comments (7)
  1. [Section IV.A] In the sentence 'one-ninth of that of DRRNN (2.5 vs. 23.9)', 'DRRNN' is a typo for 'DPRNN'.
  2. [Table III] In the U-Mamba-Net row, the numbers are formatted with stray spaces ('17 .67', '4 .4 M', '2 .5'); please fix the formatting.
  3. [Section III.C] The objective function is described as 'scale-invariant single-to-noise ratio' in two places; this should be 'scale-invariant signal-to-noise ratio'.
  4. [Section II.A and Algorithm 1] The notation is inconsistent: the input feature is defined as X in Section II.A, but Algorithm 1 uses M b as the input and output of the b-th U-Mamba block. Clarify the relationship between X, M b, and the final mask estimate.
  5. [Section II.A] The sentence 'one U-Mamba block mainly owns a U-net module and a Mamba module' should be rephrased, e.g., 'each U-Mamba block consists of a U-net module and a Mamba module'.
  6. [Abstract and Conclusion] The abstract says the model 'achieves improved performance' and the introduction says it performs 'better cross various metrics', but Table IV shows DPRNN-CMTL is superior on PESQ, CSIG, CBAK, and COVL. Please qualify these statements to reflect that the advantage is mainly on SNR-style metrics and STOI.
  7. [Figure 1] In Fig. 1b, the right panel illustrating Mamba does not label the components (Conv, FC, SSM, SiLU) with the same terminology used in the text; consider adding annotations for clarity.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: empirical comparisons are not self-referential; the only self-citation is in the DPRNN-CMTL baseline implementation, and baseline protocol is understated.

full rationale

U-Mamba-Net is an empirical architecture paper: it takes noisy-reverberant waveforms, estimates masks via alternating U-Net and Mamba blocks, and optimizes permutation-invariant SI-SNR. There is no analytic derivation from which the reported SI-SNRi improvements are forced; the numbers are measurements on a held-out simulated test set. The paper specifies disjoint training, validation, and test splits (13900/3000/3000 samples) and states that upsampling choices were checked against validation behavior ('we observe a decreased performance in the validation set during training'), so no test-set fitting is apparent. The main baselines (TasNet, SuDoRM-RF, Conv-TasNet, DPRNN) are externally published methods; the DPRNN-CMTL row is implemented by the authors using their own Enhancement Priority Pipeline from reference [3], which is a self-citation. That self-citation is a comparison-design choice, not a load-bearing premise: the CMTL baseline's performance does not presuppose U-Mamba-Net's superiority, and no equation in the paper reduces the reported gain to the EPP definition. A separate concern is that the paper never states whether all baselines in Table III were retrained and evaluated under the same 8 kHz, T60 0.2-0.6 s, WHAM!-noise simulation used for the proposed model; if the baseline numbers come from the original papers' standard Libri2mix setting, the accuracy comparison is confounded. This is an evaluation-validity issue, not circularity. The paper also openly acknowledges its weaker denoising/perceptual results relative to DPRNN-CMTL (Section IV.B, V), so it does not overclaim those metrics. Because no prediction reduces by construction to a fitted input, and no load-bearing argument is supported only by a self-citation, the circularity score is low; the two points reflect the minor self-citation in the CMTL baseline and the unstated baseline protocol rather than any circular derivation.

Assumptions & free parameters 10 free parameters · 5 assumptions · 0 invented entities

The central claim is an empirical performance comparison, so the ledger records the hand-set hyperparameters and unstated assumptions that make the comparison interpretable. No new physical or conceptual entities are introduced; the U-Mamba block is an architectural artifact rather than an invented entity.

free parameters (10)
  • Feature dimension F = 128 (default)
    Table II; ablations show SI-SNRi rises from 7.12 dB at F=64 to 8.85 dB at F=192; 128 is the chosen default.
  • Number of U-Mamba blocks R = 16
    Table II; examined 12, 16, 20; 16 chosen as the default trade-off.
  • U-Net down/up depth L = 4
    Table II; L=8 slightly hurts test performance (8.42 vs 8.50 dB SI-SNRi).
  • Upsampling method = T-Conv1D
    Table II and Table V; NN and Linear give similar test scores but worse validation behavior.
  • Encoder window/hop = 41/20 samples
    Table II; set without ablations.
  • Initial learning rate = 0.00015
    Section III.C; no schedule or decay is specified.
  • Batch size and max epochs = 4 / 120
    Section III.C; chosen by the authors.
  • Sample length and sampling rate = 3 s / 8 kHz
    Section III.C; compute load is reported per 3-s utterance.
  • Reverberation time T60 range = U(0.2, 0.6) s
    Table I; simulation setting chosen to represent typical indoor environments.
  • Train/validation/test split sizes = 13900 / 3000 / 3000
    Section III.A; fixed after simulation, not fit to the model.
assumptions (5)
  • standard math SSM state-space formulation and bilinear discretization (Eqs. 1-8) from Mamba and S4 are correct and applicable.
    The paper adopts Mamba's published equations without modification; these are treated as background from references [20] and [24].
  • domain assumption The simulated noisy-reverberant Libri2mix (WHAM! noise plus Pyroomacoustics RIRs, 8 kHz) is a valid testbed for noisy and reverberant speech separation.
    Section III.A defines the task through this specific simulation; the utility of the claimed improvement depends on this proxy being representative.
  • domain assumption SI-SNRi, SDRi, SIRi, STOI, PESQ, CSIG, CBAK, and COVL are accepted objective measures, and SI-SNR is a standard training loss.
    Evaluation in Section III.D uses these published metrics; no formal listening test is reported beyond the authors' impression in Section IV.D.
  • domain assumption The baseline results in Table III are comparable to the proposed model's results, i.e., they are measured on the same test setting.
    Section IV.A interprets Table III as a direct comparison, yet the baseline training protocol is not described. This is the paper's load-bearing but unstated premise.
  • ad hoc to paper The U-Mamba block design (U-Net then Mamba with residual connections) is a reasonable feature extraction choice.
    Section II.A and Algorithm 1 define the architecture heuristically; the design is justified only by downstream performance, with no theoretical proof.

how reviews work

0 comments
Cite this review

Pith. "Pith review of U-Mamba-Net: A highly efficient Mamba-based U-net style network for noisy and reverberant speech separation." pith.science (2026). https://pith.science/paper/WIH3ZOJT

@misc{pith2026241218217,
  author       = {Pith},
  title        = {Pith review of: U-Mamba-Net: A highly efficient Mamba-based U-net style network for noisy and reverberant speech separation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WIH3ZOJT}},
  note         = {Machine review of arXiv:2412.18217}
}
read the original abstract

The topic of speech separation involves separating mixed speech with multiple overlapping speakers into several streams, with each stream containing speech from only one speaker. Many highly effective models have emerged and proliferated rapidly over time. However, the size and computational load of these models have also increased accordingly. This is a disaster for the community, as researchers need more time and computational resources to reproduce and compare existing models. In this paper, we propose U-mamba-net: a lightweight Mamba-based U-style model for speech separation in complex environments. Mamba is a state space sequence model that incorporates feature selection capabilities. U-style network is a fully convolutional neural network whose symmetric contracting and expansive paths are able to learn multi-resolution features. In our work, Mamba serves as a feature filter, alternating with U-Net. We test the proposed model on Libri2mix. The results show that U-Mamba-Net achieves improved performance with quite low computational cost.

Figures

Figures reproduced from arXiv: 2412.18217 by the authors.

Figure 1
Figure 1. Overview of U-Mamba-Net. A. U-Mamba blocks Given the feature X ∈ R F ×T of a noisy and reverberant mixed speech, the goal of the B stacked U-mamba blocks is to estimate robust representation M ∈ R F ×T . Then, a convolutional layer helps to generate masks equal to the number of sources S. As the core part of our proposed model, one U-Mamba block mainly owns a U-net module and a Mamba module. The U-net module is comp… view at source ↗
Figure 2
Figure 2. Spectrogram of separation results. The sole spectrogram in the first [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 26 canonical work pages

  1. [3]

    A separation priority pipeline for single-channel speech separation in noisy environments,

    S. Dang, T. Matsumoto, Y . Takeuchi, and H. Kudo, “A separation priority pipeline for single-channel speech separation in noisy environments,” in IEEE Interna- tional Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2024, pp. 12 511–12 515

    work page 2024

  2. [1]

    Whamr!: Noisy and reverberant single-channel speech separation,

    M. Maciejewski, G. Wichern, E. McQuinn, and J. Le Roux, “Whamr!: Noisy and reverberant single-channel speech separation,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , IEEE, 2020, pp. 696–700

    work page 2020

  3. [2]

    Emotion awareness in multi-utterance turn for improving emotion prediction in multi-speaker conversation,

    X. Shi, X. Li, and T. Toda, “Emotion awareness in multi-utterance turn for improving emotion prediction in multi-speaker conversation,” in Proc. Interspeech, 2023, pp. 765–769

    work page 2023

  4. [4]

    Uformer: A unet based dilated complex & real dual-path conformer network for simultaneous speech enhancement and dereverberation,

    Y . Fu, Y . Liu, J. Li, et al. , “Uformer: A unet based dilated complex & real dual-path conformer network for simultaneous speech enhancement and dereverberation,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2022, pp. 7417– 7421

    work page 2022

  5. [5]

    Spec- trograms fusion with minimum difference masks esti- mation for monaural speech dereverberation,

    H. Shi, L. Wang, M. Ge, S. Li, and J. Dang, “Spec- trograms fusion with minimum difference masks esti- mation for monaural speech dereverberation,” in IEEE International Conference on Acoustics, Speech and Sig- nal Processing (ICASSP) , IEEE, 2020, pp. 7544–7548

    work page 2020

  6. [6]

    Multi-level knowledge distillation for speech emotion recognition in noisy con- ditions,

    Y . Liu, H. Sun, G. Chen, et al., “Multi-level knowledge distillation for speech emotion recognition in noisy con- ditions,” in INTERSPEECH 2023, 2023, pp. 1893–1897

    work page 2023

  7. [7]

    Mul- titalker speech separation with utterance-level permu- tation invariant training of deep recurrent neural net- works,

    M. Kolbæk, D. Yu, Z.-H. Tan, and J. Jensen, “Mul- titalker speech separation with utterance-level permu- tation invariant training of deep recurrent neural net- works,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 25, no. 10, pp. 1901–1913, 2017

    work page 1901

  8. [8]

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

    Y . Hu, Y . Liu, S. Lv, et al. , “Dccrn: Deep complex convolution recurrent network for phase-aware speech enhancement,” in Interspeech 2020 , 2020, pp. 2472– 2476

    work page 2020

Show all 31 references
  1. [9]

    Diffusion- based speech enhancement with joint generative and predictive decoders,

    H. Shi, K. Shimada, M. Hirano, et al. , “Diffusion- based speech enhancement with joint generative and predictive decoders,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2024, pp. 12 951–12 955. DOI: 10.1109/ICASSP48485. 2024.10448429

    2024

  2. [10]

    Waveform- domain speech enhancement using spectrogram encod- ing for robust speech recognition,

    H. Shi, M. Mimura, and T. Kawahara, “Waveform- domain speech enhancement using spectrogram encod- ing for robust speech recognition,” IEEE/ACM Trans- actions on Audio, Speech, and Language Processing , vol. 32, pp. 3049–3060, 2024

    2024

  3. [11]

    Dual-path rnn: Ef- ficient long sequence modeling for time-domain single- channel speech separation,

    Y . Luo, Z. Chen, and T. Yoshioka, “Dual-path rnn: Ef- ficient long sequence modeling for time-domain single- channel speech separation,” in IEEE International Con- ference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2020, pp. 46–50

    2020

  4. [12]

    Target speaker extraction with curriculum learning,

    Y . Liu, X. Liu, X. Miao, and J. Yamagishi, “Target speaker extraction with curriculum learning,” in Inter- speech 2024 , 2024, pp. 4348–4352. DOI: 10 . 21437 / Interspeech.2024-1375

    2024

  5. [13]

    A restriction training recipe for speech separation on sparsely mixed speech,

    S. Dang, T. Matsumoto, H. Kudo, and Y . Takeuchi, “A restriction training recipe for speech separation on sparsely mixed speech,” in Neural Information Pro- cessing: 28th International Conference, ICONIP 2021, Sanur , Bali, Indonesia, December 8–12, 2021, Proceed- ings, Part V...

    2021

  6. [14]

    Attention is all you need in speech separation,

    C. Subakan, M. Ravanelli, S. Cornell, M. Bronzi, and J. Zhong, “Attention is all you need in speech separation,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , IEEE, 2021, pp. 21– 25

    2021

  7. [15]

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

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

    2019

  8. [16]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, et al., “Attention is all you need,” Advances in neural information process- ing systems , vol. 30, 2017

    2017

  9. [17]

    U-net: Convo- lutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convo- lutional networks for biomedical image segmentation,” in Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international confer- ence, Munich, Germany, October 5-9, 2015, proceed- ings, part III...

    2015

  10. [18]

    D ´efossez, N

    A. D ´efossez, N. Usunier, L. Bottou, and F. Bach, Demucs: Deep extractor for music sources with ex- tra unlabeled data remixed , 2019. arXiv: 1909.01174 [cs.SD]

    2019 arXiv

  11. [19]

    Sudo rm -rf: Efficient networks for universal audio source separa- tion,

    E. Tzinis, Z. Wang, and P. Smaragdis, “Sudo rm -rf: Efficient networks for universal audio source separa- tion,” in 2020 IEEE 30th International Workshop on Machine Learning for Signal Processing (MLSP) , IEEE, Sep. 2020

    2020

  12. [20]

    Mamba: Linear-time sequence modeling with selective state spaces,

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

    2023 arXiv

  13. [21]

    How to train your hippo: State space models with generalized orthogonal basis projections,

    A. Gu, I. Johnson, A. Timalsina, A. Rudra, and C. R ´e, “How to train your hippo: State space models with generalized orthogonal basis projections,” arXiv preprint arXiv:2206.12037, 2022

    2022 arXiv

  14. [22]

    Cosentino, M

    J. Cosentino, M. Pariente, S. Cornell, A. Deleforge, and E. Vincent, Librimix: An open-source dataset for generalizable speech separation , 2020. arXiv: 2005 . 11262 [eess.AS]. 5

    2020

  15. [23]

    Us- ing semi-supervised learning for monaural time-domain speech separation with a self-supervised learning-based si-snr estimator,

    S. Dang, T. Matsumoto, Y . Takeuchi, and H. Kudo, “Us- ing semi-supervised learning for monaural time-domain speech separation with a self-supervised learning-based si-snr estimator,” in Interspeech 2023. ISCA, 2023

    2023

  16. [24]

    Efficiently modeling long sequences with structured state spaces,

    A. Gu, K. Goel, and C. R ´e, “Efficiently modeling long sequences with structured state spaces,” arXiv preprint arXiv:2111.00396, 2021

    2021 arXiv

  17. [25]

    WHAM!: Extending speech separation to noisy environments,

    G. Wichern, J. Antognini, M. Flynn, et al. , “WHAM!: Extending speech separation to noisy environments,” in 20th Annual Conference of the International Speech Communication Association, Interspeech 2019, Graz, Austria, September 15-19, 2019 , ISCA, 2019, pp. 1368– 1372

    2019

  18. [26]

    Pyrooma- coustics: A python package for audio room simulation and array processing algorithms,

    R. Scheibler, E. Bezzam, and I. Dokmani ´c, “Pyrooma- coustics: A python package for audio room simulation and array processing algorithms,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2018, pp. 351–355

    2018

  19. [27]

    SDR–half-baked or well done?

    J. Le Roux, S. Wisdom, H. Erdogan, and J. R. Hershey, “SDR–half-baked or well done?” In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2019, pp. 626–630

    2019

  20. [28]

    Performance measurement in blind audio source separation,

    E. Vincent, R. Gribonval, and C. F ´evotte, “Performance measurement in blind audio source separation,” IEEE Transactions on Audio, Speech, and Language Process- ing, vol. 14, no. 4, pp. 1462–1469, 2006

    2006

  21. [29]

    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,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , I...

    2001

  22. [30]

    Evaluation of objective quality measures for speech enhancement,

    Y . Hu and P. C. Loizou, “Evaluation of objective quality measures for speech enhancement,” IEEE Transactions on Audio, Speech, and Language Processing , vol. 16, no. 1, pp. 229–238, 2007

    2007

  23. [31]

    Ptflops: A flops counting tool for neu- ral networks in pytorch framework

    V . Sovrasov. “Ptflops: A flops counting tool for neu- ral networks in pytorch framework.” (2024), [Online]. Available: https://github.com/sovrasov/flops- counter. pytorch. 6

    2024

Pith tools

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