Pith. sign in

REVIEW 2 major objections 52 references

TF-MoE: Time-Frequency Mixture-of-Experts for Efficient Speech Separation

T0 review · 2 major / 0 minor · reviewed 2026-07-01 · grok-4.3

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

Pith's one-line read TF-MoE adds dynamic time and frequency expert selection to raise speech separation quality at fixed low compute cost.

desk verdict TF-MoE claims a 3.8 dB SDR gain over BSRNN at matched 4.1 GMACs/s via time-frequency MoE on Conformer, but the efficiency claim needs full methods to confirm. read the letter →

arxiv 2606.29575 v2 pith:HETK53KG submitted 2026-06-28 cs.SD cs.AI

Qinzhe Hu , Chenda Li , Wangyou Zhang , Shujie Liu , Yan Lu , Yanmin Qian This is my paper
classification cs.SDcs.AI
keywords speechseparationmixtureofexpertstime-frequencymodelingefficientneuralnetworksConformerbackboneedgedevicedeployment
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 introduces TF-MoE as a sparse Mixture-of-Experts architecture for speech separation that alternates between time-wise and frequency-wise modules, each choosing experts dynamically per frame or mel band. This design is placed on a mel-band-splitting Conformer backbone to grow model capacity while keeping inference cost nearly unchanged. The central result is that the approach delivers higher separation scores than prior compact models under the same computation budget. If the claim holds, edge devices could run higher-quality separation without extra hardware resources.

What carries the argument

Alternating time-wise and frequency-wise MoE modules with per-frame or per-mel-band dynamic expert selection.

What would settle it

Run the model on target edge hardware and measure end-to-end latency and power; if the measured cost exceeds the reported 4.1 GMACs/s or the SDR gain disappears when overhead is included, the claim does not hold.

Watch

Extended reading notes

Core claim

TF-MoE introduces alternating time-wise and frequency-wise MoE modules that perform dynamic expert selection per frame or mel band on a mel-band-splitting Conformer backbone, raising separation performance under strict computation limits while adding almost no inference cost.

Load-bearing premise

Dynamic expert selection in the time and frequency modules adds negligible extra computation beyond the stated GMACs figure and produces real specialization instead of duplicated work.

Editorial extensions

If this is right

  • TF-MoE outperforms BSRNN by 3.8 dB SDR on Libri2Mix at comparable 4.1 GMACs/s inference cost.
  • Model capacity grows without a proportional rise in inference cost for speech separation tasks.
  • The architecture is positioned for deployment on edge devices that face tight compute budgets.
  • Performance gains appear consistently when computation cost is held fixed.

Reading between the lines

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

  • The same alternating selection pattern could be tested on other audio tasks such as enhancement or diarization that also split time and frequency processing.
  • If the experts truly specialize, one could inspect activation patterns to see whether time experts handle transients and frequency experts handle harmonics.
  • Combining the MoE structure with further pruning or quantization might push the operating point even lower while preserving the reported gains.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 0 minor

Summary. The paper proposes TF-MoE, a sparse Mixture-of-Experts architecture for speech separation built on a mel-band-splitting Conformer backbone. It introduces alternating time-wise and frequency-wise MoE modules that perform dynamic expert selection per frame or mel band. The central empirical claim is that this yields consistent gains under compute constraints, specifically +3.8 dB SDR improvement over BSRNN on Libri2Mix at a comparable inference cost of 4.1 GMACs/s, positioning the model for edge deployment.

Significance. If the efficiency and performance claims are substantiated with full cost accounting and specialization evidence, the work would offer a practical route to higher-capacity speech separation models without proportional compute growth, which is relevant for on-device audio processing.

major comments (2)
  1. [Abstract] Abstract: the headline result (+3.8 dB SDR at 4.1 GMACs/s) rests on the unverified assertion that alternating time/frequency MoE modules add negligible inference cost beyond the stated GMACs figure. No derivation or breakdown of MACs for the gating networks, expert loading, or dynamic routing is supplied, so it is impossible to confirm that the comparison to BSRNN remains valid once routing overhead is included.
  2. [Abstract] Abstract: the claim of 'genuine specialization' via per-frame/per-mel-band expert selection lacks supporting statistics (e.g., expert utilization histograms or activation overlap metrics). Without these, the efficiency advantage could be an artifact of redundant computation paths rather than sparse capacity increase.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the computational cost accounting and evidence of specialization. We address each major comment below and will revise the manuscript to incorporate the requested details.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the headline result (+3.8 dB SDR at 4.1 GMACs/s) rests on the unverified assertion that alternating time/frequency MoE modules add negligible inference cost beyond the stated GMACs figure. No derivation or breakdown of MACs for the gating networks, expert loading, or dynamic routing is supplied, so it is impossible to confirm that the comparison to BSRNN remains valid once routing overhead is included.

    Authors: We agree that an explicit breakdown strengthens the efficiency claim. The reported 4.1 GMACs/s was obtained via standard FLOPs profiling of the complete architecture (including gating and routing) on the target hardware, but the abstract and main text do not provide the component-wise derivation. We will add a dedicated paragraph and supplementary table in the revised manuscript that decomposes the MACs for the time-wise and frequency-wise MoE modules, gating networks, and routing overhead to allow direct verification against BSRNN. revision: yes

  2. Referee: [Abstract] Abstract: the claim of 'genuine specialization' via per-frame/per-mel-band expert selection lacks supporting statistics (e.g., expert utilization histograms or activation overlap metrics). Without these, the efficiency advantage could be an artifact of redundant computation paths rather than sparse capacity increase.

    Authors: We acknowledge that the current manuscript does not include quantitative evidence of expert specialization. To demonstrate that the per-frame and per-mel-band routing produces non-redundant expert usage, we will add expert utilization histograms, load-balancing statistics, and activation overlap metrics (e.g., average pairwise expert co-activation) to the experiments section of the revised paper. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: purely empirical architecture proposal with reported experimental outcomes

full rationale

The paper introduces TF-MoE as a sparse MoE architecture on a mel-band-splitting Conformer backbone and reports measured SDR gains and GMACs/s figures from experiments on Libri2Mix. No equations, first-principles derivations, fitted parameters renamed as predictions, or self-citation chains appear in the provided text. The +3.8 dB claim is presented as an observed experimental result rather than a quantity forced by construction from model definitions or prior self-citations. The derivation chain is therefore empty and self-contained.

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

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the method implicitly assumes that sparse expert routing incurs no hidden compute cost and that mel-band splitting is compatible with the MoE modules.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TF-MoE: Time-Frequency Mixture-of-Experts for Efficient Speech Separation." pith.science (2026). https://pith.science/paper/HETK53KG

@misc{pith2026260629575,
  author       = {Pith},
  title        = {Pith review of: TF-MoE: Time-Frequency Mixture-of-Experts for Efficient Speech Separation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HETK53KG}},
  note         = {Machine review of arXiv:2606.29575}
}
read the original abstract

Recent advances in speech separation (SS) have led to compact front-end models with small parameter sizes, yet their high computational cost remains a major barrier for deployment on edge devices. To address this, we propose TF-MoE, a sparse Mixture-of-Experts (MoE) framework that enhances model capacity with almost no increase in inference cost. Our method introduces dynamic expert specialization in time and frequency dimensions through alternating time-wise and frequency-wise MoE modules, each dynamically selecting experts per frame or mel band. Built upon a mel-band-splitting Conformer backbone, TF-MoE achieves strong performance on SS tasks under low-compute settings. Experimental results demonstrate that TF-MoE consistently improves separation performance under computation cost constraints, outperforming BSRNN by +3.8 dB SDR on Libri2Mix with comparable 4.1 GMACs/s inference cost. This positions TF-MoE as a promising candidate for edge-device deployment.

Figures

Figures reproduced from arXiv: 2606.29575 by the authors.

Figure 1
Figure 1. Overview of the proposed TF-MoE framework. (A) The TF-Conformer backbone; (B) The proposed TF-MoE model; (C) The MoE Conformer block, where standard feed-forward modules are replaced by MoE FFNs; (D.1): The standard feed-forward module; (D.2) The MoE feed-forward module with top-J sparse routing. 2. Methodology 2.1. Backbone Model We first introduce a new SS model, TF-Conformer, as our back￾bone, which is inspired b… view at source ↗
Figure 2
Figure 2. Visualization of the expert routing in the first Feed Forward module of the fifth TF-MoE block. The top spec￾trogram displays the time-frequency representation of the in￾put signal comprising female-only (frames 0-100), mixed (100- 200), and male-only (200-300) segments. The lower bar shows the routing decisions of T-MoE experts in the T-module along the time axis, while the right bar shows the routing decisions of … view at source ↗

Review history (2 revisions) →

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

52 extracted references · 52 canonical work pages

  1. [1]

    Introduction Speech separation (SS) has made great progress with deep learning [1, 2, 3, 4, 5, 6, 7, 8], emerging as a critical enabler for a wide range of real-world applications. Driven by growing pri- vacy concerns [9], the demand for low-latency interaction [10], and the need for offline capability in network-unavailable envi- ronments [11, 12], there...

    work page

  2. [2]

    We propose a competitive mel-band-splitting Conformer backbone that balances performance and efficiency, achieving a +2.5 dB SDR improvement over BSRNN on Libri2Mix at com- parable computational cost

    work page

  3. [3]

    It outperforms Conformer backbone by +1.3 dB SDR on Libri2Mix

    We propose TF-MoE, a sparse Mixture-of-Experts framework that replaces the feed-forward modules in both the time and frequency Conformer blocks with sparsely-gated ex- pert layers, scaling model capacity without increasing compu- tational cost. It outperforms Conformer backbone by +1.3 dB SDR on Libri2Mix

    work page

  4. [4]

    TF-MoE: Time-Frequency Mixture-of-Experts for Efficient Speech Separation

    Comprehensive ablation studies validate the complemen- tary nature of sparse MoE routing and Conformer on both tem- poral and frequency dimensions. Furthermore, visual analysis of the gating policies reveals an explicit structural specialization on different acoustic patterns. arXiv:2606.29575v2 [cs.SD] 30 Jun 2026 Feed Forward / MoE Multi-head Self Atten...

    work page Pith review arXiv 2026

  5. [5]

    Methodology 2.1. Backbone Model We first introduce a new SS model, TF-Conformer, as our back- bone, which is inspired by the band-split RNN (BSRNN) [6] with two key modifications: (i) the manually designed sub- band splitting is replaced with mel-scale [30] band splitting, and (ii) the RNN sequence modeling modules are replaced with Conformer [31] blocks....

    work page

  6. [6]

    Experimental Setup Datasets & Configuration:Experiments are conducted on the Libri2Mix (16kHz, min) dataset from LibriMix [33]

    Experiments 3.1. Experimental Setup Datasets & Configuration:Experiments are conducted on the Libri2Mix (16kHz, min) dataset from LibriMix [33]. The input spectrogram is extracted using a 32 ms Hanning window with an 8 ms shift, and split intoK=80 mel-scale [30] sub-bands. Unless otherwise stated, models use hidden dimensionN= 32 Table 1:Separation result...

    work page

  7. [7]

    free lunch

    Conclusion We proposed TF-MoE, a highly efficient time-frequency Mixture-of-Experts framework, to overcome the capacity- computation bottleneck in the state-of-the-art speech separa- tion. By integrating sparse expert routing into a dual-path Con- former backbone, our model successfully converts newly added parameters into separation performance almost wi...

    work page

  8. [8]

    2021ZD0201500, in part by the Na- tional Natural Science Foundation of China under Grant No

    Acknowledgments This work was supported in part by the China STI 2030–Major Projects under Grant No. 2021ZD0201500, in part by the Na- tional Natural Science Foundation of China under Grant No. U25A20409, and in part by the SJTU Med-X (Medicine & Engineering) Translational Research Grant under Grant No. YG2025LC09

    work page

Show all 52 references
  1. [9]

    Generative AI Use Disclosure During the preparation of this work, the author(s) used genera- tive AI tools (e.g., ChatGPT, GLM, Gemini) for language pol- ishing and grammar checking to improve the readability of the manuscript. All scientific content, methodology design, and e...

  2. [10]

    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 Au- dio, Speech, and Language Processing, vol. 26, no. 10, pp. 1702– 1726, 2018

    2018

  3. [11]

    Permutation invari- ant training of deep models for speaker-independent multi-talker speech separation,

    D. Yu, M. Kolbæk, Z. Tan, and J. Jensen, “Permutation invari- ant training of deep models for speaker-independent multi-talker speech separation,” Mar. 2017, pp. 241–245

    2017

  4. [12]

    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,” vol. 27, no. 8, pp. 1256–1266, Aug. 2019

    2019

  5. [13]

    Speech Enhancement with Score-Based Generative Models in the Complex STFT Do- main,

    S. Welker, J. Richter, and T. Gerkmann, “Speech Enhancement with Score-Based Generative Models in the Complex STFT Do- main,” inInterspeech 2022. ISCA, Sep. 2022, pp. 2928–2932

    2022

  6. [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. Watan- abe, “Tf-gridnet: Making time-frequency domain models great again for monaural speaker separation,” inICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023, pp. 1–5

    2023

  7. [15]

    Efficient Monaural Speech Enhancement with Universal Sample Rate Band-Split RNN,

    J. Yu and Y . Luo, “Efficient Monaural Speech Enhancement with Universal Sample Rate Band-Split RNN,” Jun. 2023, pp. 1–5

    2023

  8. [16]

    URGENT Challenge: Universality, Robustness, and Generalizability For Speech Enhancement,

    W. Zhang, R. Scheibler, K. Saijo, S. Cornell, C. Li, Z. Ni, A. Kumar, J. Pirklbauer, M. Sach, S. Watanabe, T. Fingscheidt, and Y . Qian, “URGENT Challenge: Universality, Robustness, and Generalizability For Speech Enhancement,” 2024, pp. 4868– 4872

    2024

  9. [17]

    Interspeech 2025 URGENT Speech Enhancement Challenge,

    K. Saijo, W. Zhang, S. Cornell, R. Scheibler, C. Li, Z. Ni, A. Ku- mar, M. Sach, Y . Fu, W. Wang, T. Fingscheidt, and S. Watanabe, “Interspeech 2025 URGENT Speech Enhancement Challenge,” in Interspeech 2025, 2025, pp. 858–862

    2025

  10. [18]

    The voiceprivacy 2024 challenge evaluation plan,

    N. Tomashenko, X. Miao, P. Champion, S. Meyer, X. Wang, E. Vincent, M. Panariello, N. Evans, J. Yamagishi, and M. Todisco, “The voiceprivacy 2024 challenge evaluation plan,” arXiv preprint arXiv:2404.02677, 2024

    2024

  11. [19]

    Icassp 2023 deep noise suppression challenge,

    H. Dubey, A. Aazami, V . Gopal, B. Naderi, S. Braun, R. Cutler, A. Ju, M. Zohourian, M. Tang, M. Golestanehet al., “Icassp 2023 deep noise suppression challenge,”IEEE Open Journal of Signal Processing, vol. 5, pp. 725–737, 2024

    2023

  12. [20]

    Tinylstms: Effi- cient neural speech enhancement for hearing aids,

    I. Fedorov, M. Stamenovic, C. Jensen, L.-C. Yang, A. Mandell, Y . Gan, M. Mattina, and P. N. Whatmough, “Tinylstms: Effi- cient neural speech enhancement for hearing aids,”arXiv preprint arXiv:2005.11138, 2020

    2005

  13. [21]

    Low latency speech enhancement for hearing aids using deep filtering,

    H. Schr ¨oter, T. Rosenkranz, A.-N. Escalante-B, and A. Maier, “Low latency speech enhancement for hearing aids using deep filtering,”IEEE/ACM Transactions on Audio, Speech, and Lan- guage Processing, vol. 30, pp. 2716–2728, 2022

    2022

  14. [22]

    Dual-Path RNN: Effi- cient Long Sequence Modeling for Time-Domain Single-Channel Speech Separation,

    Y . Luo, Z. Chen, and T. Yoshioka, “Dual-Path RNN: Effi- cient Long Sequence Modeling for Time-Domain Single-Channel Speech Separation,” Barcelona, May 2020, pp. 46–50

    2020

  15. [23]

    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,” Jun. 2021, pp. 21–25

    2021

  16. [24]

    Skim: Skipping Mem- ory Lstm for Low-Latency Real-Time Continuous Speech Sepa- ration,

    C. Li, L. Yang, W. Wang, and Y . Qian, “Skim: Skipping Mem- ory Lstm for Low-Latency Real-Time Continuous Speech Sepa- ration,” Singapore, May 2022, pp. 681–685

    2022

  17. [25]

    Tiger: Time-frequency in- terleaved gain extraction and reconstruction for efficient speech separation,

    M. Xu, K. Li, G. Chen, and X. Hu, “Tiger: Time-frequency in- terleaved gain extraction and reconstruction for efficient speech separation,”arXiv preprint arXiv:2410.01469, 2024

    2024

  18. [26]

    Stack Less, Repeat More: A Block Reusing Approach for Progressive Speech En- hancement,

    J. Kim, U.-H. Shin, J. Ko, and H.-M. Park, “Stack Less, Repeat More: A Block Reusing Approach for Progressive Speech En- hancement,” May 2025

    2025

  19. [27]

    Beyond Performance Plateaus: A Comprehensive Study on Scal- ability in Speech Enhancement,

    W. Zhang, K. Saijo, J.-w. Jung, C. Li, S. Watanabe, and Y . Qian, “Beyond Performance Plateaus: A Comprehensive Study on Scal- ability in Speech Enhancement,” inProc. Interspeech 2024, 2024, pp. 1740–1744

    2024

  20. [28]

    Predictive Skim: Contrastive Pre- dictive Coding for Low-Latency Online Speech Separation,

    C. Li, Y . Wu, and Y . Qian, “Predictive Skim: Contrastive Pre- dictive Coding for Low-Latency Online Speech Separation,” Jun. 2023, pp. 1–5

    2023

  21. [29]

    CheapNET: Improving Light- weight speech enhancement network by projected loss function,

    K. Tan, B. Dai, J. Li, and W. Mao, “CheapNET: Improving Light- weight speech enhancement network by projected loss function,” Nov. 2023

    2023

  22. [30]

    TF-SkiMNet: Speech En- hancement Based on Inplace Modeling and Skipping Memory in Time-Frequency Domain,

    Z. Li, S. He, J. Bai, and X. Zhang, “TF-SkiMNet: Speech En- hancement Based on Inplace Modeling and Skipping Memory in Time-Frequency Domain,” 2025, pp. 5143–5147

    2025

  23. [31]

    A Comprehensive Survey of Mixture-of- Experts: Algorithms, Theory, and Applications,

    S. Mu and S. Lin, “A Comprehensive Survey of Mixture-of- Experts: Algorithms, Theory, and Applications,” Apr. 2025

    2025

  24. [32]

    Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer,

    N. Shazeer, A. Mirhoseini, K. Maziarz, A. Davis, Q. Le, G. Hin- ton, and J. Dean, “Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer,” Jan. 2017

    2017

  25. [33]

    Building a great multi-lingual teacher with sparsely-gated mixture of experts for speech recognition,

    K. Kumatani, R. Gmyr, F. C. Salinas, L. Liu, W. Zuo, D. Patel, E. Sun, and Y . Shi, “Building a great multi-lingual teacher with sparsely-gated mixture of experts for speech recognition,” Jan. 2022

    2022

  26. [34]

    GLaM: Efficient Scaling of Language Models with Mixture-of-Experts,

    N. Du, Y . Huang, A. M. Dai, S. Tong, D. Lepikhin, Y . Xu, M. Krikun, Y . Zhou, A. W. Yu, O. Firat, B. Zoph, L. Fedus, M. P. Bosma, Z. Zhou, T. Wang, E. Wang, K. Webster, M. Pellat, K. Robinson, K. Meier-Hellstern, T. Duke, L. Dixon, K. Zhang, Q. Le, Y . Wu, Z. Chen, and C. Cu...

    2022

  27. [35]

    DeepSeekMoE: Towards Ulti- mate Expert Specialization in Mixture-of-Experts Language Mod- els,

    D. Dai, C. Deng, C. Zhao, R. X. Xu, H. Gao, D. Chen, J. Li, W. Zeng, X. Yu, Y . Wu, Z. Xie, Y . K. Li, P. Huang, F. Luo, C. Ruan, Z. Sui, and W. Liang, “DeepSeekMoE: Towards Ulti- mate Expert Specialization in Mixture-of-Experts Language Mod- els,” Jan. 2024

    2024

  28. [36]

    CLIP-MoE: Towards Building Mixture of Experts for CLIP with Diversified Multiplet Upcycling,

    J. Zhang, X. Qu, T. Zhu, and Y . Cheng, “CLIP-MoE: Towards Building Mixture of Experts for CLIP with Diversified Multiplet Upcycling,” May 2025

    2025

  29. [37]

    Speech Enhance- ment using a Deep Mixture of Experts,

    S. E. Chazan, J. Goldberger, and S. Gannot, “Speech Enhance- ment using a Deep Mixture of Experts,” Mar. 2017

    2017

  30. [38]

    Handling Trade-Offs in Speech Separation with Sparsely-Gated Mixture of Experts,

    X. Wang, Z. Chen, Y . Shi, J. Wu, N. Kanda, and T. Yoshioka, “Handling Trade-Offs in Speech Separation with Sparsely-Gated Mixture of Experts,” May 2023

    2023

  31. [39]

    A Scale for the Measurement of the Psychological Magnitude Pitch,

    J. V olkmann, S. S. Stevens, and E. B. Newman, “A Scale for the Measurement of the Psychological Magnitude Pitch,”The Journal of the Acoustical Society of America, vol. 8, no. 3 Supplement, p. 208, Jan. 1937

    1937

  32. [40]

    Conformer: Convolution-augmented Transformer for Speech Recognition,

    A. Gulati, J. Qin, C.-C. Chiu, N. Parmar, Y . Zhang, J. Yu, W. Han, S. Wang, Z. Zhang, Y . Wu, and R. Pang, “Conformer: Convolution-augmented Transformer for Speech Recognition,” 2020, pp. 5036–5040

    2020

  33. [41]

    Searching for activa- tion functions,

    P. Ramachandran, B. Zoph, and Q. V . Le, “Searching for activa- tion functions,”arXiv preprint arXiv:1710.05941, 2017

    2017 arXiv

  34. [42]

    Librimix: An open-source dataset for generalizable speech separation,

    J. Cosentino, M. Pariente, S. Cornell, A. Deleforge, and E. Vin- cent, “Librimix: An open-source dataset for generalizable speech separation,”arXiv, 2020

    2020

  35. [43]

    Decoupled weight decay regulariza- tion,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regulariza- tion,”arXiv preprint arXiv:1711.05101, 2017

    2017 arXiv

  36. [44]

    Sgdr: Stochastic gradient descent with warm restarts,

    ——, “Sgdr: Stochastic gradient descent with warm restarts,” arXiv preprint arXiv:1608.03983, 2016

    2016 arXiv

  37. [45]

    BSS EV AL Toolbox User Guide – Revision 2.0,

    C. F ´evotte, R. Gribonval, and E. Vincent, “BSS EV AL Toolbox User Guide – Revision 2.0,” Report, 2005

    2005

  38. [46]

    SDR – Half-baked or Well Done?

    J. L. Roux, S. Wisdom, H. Erdogan, and J. R. Hershey, “SDR – Half-baked or Well Done?” May 2019, pp. 626–630

    2019

  39. [47]

    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 2001 IEEE international conference on acoustics, speech, and signal processing. Proceed...

    2001

  40. [48]

    An Algo- rithm for Intelligibility Prediction of Time–Frequency Weighted Noisy Speech,

    C. H. Taal, R. C. Hendriks, R. Heusdens, and J. Jensen, “An Algo- rithm for Intelligibility Prediction of Time–Frequency Weighted Noisy Speech,” vol. 19, no. 7, pp. 2125–2136, Sep. 2011

    2011

  41. [49]

    SPMamba: State-space model is all you need in speech separation,

    K. Li, G. Chen, R. Yang, and X. Hu, “SPMamba: State-space model is all you need in speech separation,” Sep. 2024

    2024

  42. [50]

    Speech separation using an asynchronous fully recurrent convolutional neural network,

    X. Hu, K. Li, W. Zhang, Y . Luo, J.-M. Lemercier, and T. Gerk- mann, “Speech separation using an asynchronous fully recurrent convolutional neural network,”Advances in Neural Information Processing Systems, vol. 34, pp. 22 509–22 522, 2021

    2021

  43. [51]

    An efficient encoder-decoder ar- chitecture with top-down attention for speech separation,

    K. Li, R. Yang, and X. Hu, “An efficient encoder-decoder ar- chitecture with top-down attention for speech separation,”arXiv preprint arXiv:2209.15200, 2022

    2022

  44. [52]

    Sudo rm-rf: Efficient net- works for universal audio source separation,

    E. Tzinis, Z. Wang, and P. Smaragdis, “Sudo rm-rf: Efficient net- works for universal audio source separation,” in2020 IEEE 30th International Workshop on Machine Learning for Signal Process- ing (MLSP). IEEE, 2020, pp. 1–6

    2020

Pith tools

Reviewed July 1, 2026 · model on record in the stance chip above.