Pith. sign in

REVIEW 4 major objections 5 minor 44 references

Probabilistic Permutation Invariant Training for Speech Separation

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Probabilistic PIT replaces the hard permutation pick with a soft minimum and improves speech separation.

desk verdict A clean probabilistic re-derivation of PIT as soft-minimum training, with real but currently under-specified empirical support. read the letter →

arxiv 1908.01768 v1 pith:FUT5DYSX submitted 2019-08-04 eess.AS cs.LGcs.SDstat.ML

Midia Yousefi , Soheil Khorram , John H.L. Hansen This is my paper · ORCID
classification eess.AScs.LGcs.SDstat.ML
keywords speechseparationpermutationinvarianttrainingprobabilisticPITsoft-minimumambiguityLSTMsourcelog-likelihood
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 argues that Permutation Invariant Training (PIT) is overly confident when it forces the network to update from the single output-label assignment with the smallest separation error, especially early in training when the network's outputs are unreliable and the costs of different permutations are close. It proposes Probabilistic PIT (Prob-PIT), which treats the output-label permutation as a discrete latent variable with a uniform prior and models the estimation error as Gaussian, leading to a log-likelihood objective that replaces the minimum over permutations with a soft-minimum controlled by a smoothing factor $\gamma$. On two-talker mixtures from TIMIT and GRID, Prob-PIT beats PIT in Signal to Distortion Ratio and Signal to Interference Ratio for every tested $\gamma$, with pairwise t-tests giving $p<0.01$ for a range of $\gamma$ values. If the claim holds, speech separation networks can be trained better simply by softening the permutation selection, without any architectural change.

What carries the argument

The load-bearing object is the soft-minimum of permutation costs, Eq. (5): $Q(\theta) = -g(Z_{\min},\theta) + \gamma \log\left(1 + \sum_{Z\neq Z_{\min}} \exp\left(\frac{g(Z_{\min},\theta)-g(Z,\theta)}{\gamma}\right)\right)$. This expression is derived from a generative model $X = Z(G(Y,\theta)) + \epsilon$ with $\epsilon \sim \mathcal{N}(0,\sigma^2 I)$ and a uniform prior over $Z$; it is the log-likelihood of the clean spectra. The smoothing factor $\gamma = 2\sigma^2$ controls how much the objective attends to non-minimal permutations: $\gamma=0$ reduces to PIT, and larger $\gamma$ weights all permutations more equally. This single replacement is what changes the training dynamics, yielding the reported gains.

What would settle it

Train the same two-talker LSTM with a warm-start initialization that already separates speech before training (e.g., pretrained on clean-target reconstruction), then compare Prob-PIT against PIT from that start; if the SDR/SIR advantage disappears or reverses, the paper's diagnosis that early unreliable outputs drive the gain is wrong. Alternatively, measure the distribution of $|\text{cost}_1 - \text{cost}_2|$ across the first epoch on a new dataset; the argument requires these differences to be small often enough that hard minimums are frequently wrong.

Watch

Extended reading notes

Core claim

The central claim is that the hard minimum in PIT is a suboptimal training rule, and that maximizing a log-likelihood in which the permutation is a latent variable yields a better objective. Concretely, Prob-PIT defines $g(Z,\theta) = \|X - Z(G(Y,\theta))\|^2$ as the separation error of permutation $Z$, and maximizes $Q(\theta) = -g(Z_{\min},\theta) + \gamma \log\left(1 + \sum_{Z\neq Z_{\min}} \exp\left(\frac{g(Z_{\min},\theta)-g(Z,\theta)}{\gamma}\right)\right)$, where $Z_{\min}$ is the best permutation and $\gamma = 2\sigma^2$ is the variance parameter of the Gaussian error model. Setting $\gamma=0$ recovers standard PIT. In two-talker LSTM experiments on TIMIT and GRID, Prob-PIT outperforms PIT in SDR and SIR for all $\gamma$ values tested, with best results at $\gamma=32$ (TIMIT) and $\gamma=45$ (GRID), and the t-test results are significant at $p<0.01$ for $1<\gamma$ on GRID and $8<\gamma<40$ on TIMIT.

Load-bearing premise

The derivation assumes the estimation error $\epsilon$ is independent Gaussian with a single scalar variance $\sigma^2$; if the error is correlated, non-Gaussian, or input-dependent, the soft-minimum weights no longer capture the true uncertainty, and the claimed benefit may not transfer to other datasets or architectures.

Editorial extensions

If this is right

  • Any network trained with PIT can switch to Prob-PIT by replacing the minimum over permutations with the soft-minimum loss; no architectural change is required.
  • The smoothing parameter $\gamma$ gives practitioners a direct knob: $\gamma=0$ is PIT, while larger $\gamma$ places more weight on all permutations during training.
  • The reported gains are consistent across two corpora and both evaluation metrics, with lower result variance for reasonable $\gamma$, suggesting a smoother optimization landscape and more stable training.
  • The likelihood formulation extends in principle to any number of sources $S$, summing over all $S!$ permutations, though the paper only evaluates the two-talker case.

Reading between the lines

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

  • A testable prediction follows from the paper's diagnosis: if PIT is overconfident mainly because early outputs are unreliable, then warm-starting Prob-PIT from a pretrained PIT model should shrink or eliminate the gap; this experiment is not in the paper.
  • The same latent-permutation soft-minimum device could be applied to other problems with output-order ambiguity, such as multi-speaker speech recognition or speaker diarization, where labeling is also a discrete choice.
  • The Gaussian scalar-variance assumption is a natural next target: allowing $\sigma^2$ to be predicted per frame or per utterance could make the soft-minimum weights adapt to local uncertainty and may improve the gains further.
  • For more than two sources, the $S!$ permutation sum grows quickly; at $S=3$ there are six terms (still cheap), but larger $S$ would require sampling or approximations, which the paper does not address.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes Probabilistic Permutation Invariant Training (Prob-PIT) for single-microphone speech separation. The method treats the output-label permutation as a discrete latent variable with a uniform prior, assumes Gaussian estimation error with variance sigma^2, and derives a log-likelihood that is a soft-minimum over all permutation costs (Eq. 4, Eq. 5). Training then maximizes this soft-minimum objective instead of the hard minimum used in conventional PIT. The authors motivate the approach with a preliminary experiment showing that the two permutation costs are close in early training, and they report experiments on two-speaker separation claiming that Prob-PIT significantly outperforms PIT in SDR and SIR on TIMIT and GRID datasets, with gamma tuned per dataset and significance assessed by pairwise t-tests.

Significance. If the empirical claims hold, Prob-PIT is an elegant and practically attractive contribution: it gives a principled probabilistic interpretation to the temperature parameter of a soft-minimum and can be implemented by replacing the min in PIT with a soft-min, which is a small code change. The derivation is transparent and the connection between the generative model and the training objective is clearly stated. The preliminary visualization of overlapping permutation costs in the first epoch is a useful motivational result. However, the experimental validation has several load-bearing weaknesses concerning how the smoothing parameter gamma is selected and how the loss scaling affects training schedules; these must be resolved before the claimed significant improvements can be accepted.

major comments (4)
  1. [Sec. 4.1, Fig. 3] The gamma selection procedure is not described as a validation-set search. The text says that values are 'chosen until a decrease in the performance is observed' and then reports the best gamma (45 for GRID, 32 for TIMIT) directly from Fig. 3, with no statement that a held-out validation set was used. If Fig. 3 shows test-set performance, the t-tests at those best-gamma values are selection artifacts and the reported p-values are not valid. Please specify whether gamma was tuned on the validation set and, if so, report the selected values separately from the test-set curves.
  2. [Sec. 3, Eq. 5] Equation (5) is not the numerically stable form of Eq. (4); it is gamma times the log-sum-exp term of Eq. (4), up to a constant. Maximizing it is equivalent to maximizing Eq. (4) for a fixed gamma, but the absolute loss value scales with gamma. Since Section 4 states that the learning rate is reduced when the loss improvement is less than 0.003, the early-stopping/scheduler criterion is not scale-invariant and therefore behaves differently for each gamma and for the gamma=0 PIT baseline. This makes the training schedules across conditions incomparable. Please use the unscaled log-likelihood in Eq. (4) for scheduling, or state a scale-invariant scheduling rule.
  3. [Sec. 4.1, statistical tests] The pairwise t-tests are based on five experiments per gamma and are reported only for selected gamma ranges after inspecting Fig. 3 ('1 < gamma for GRID' and '8 < gamma < 40 for TIMIT'). With n=5, the normality assumption is untestable and the selection of ranges post hoc inflates the Type I error. Please report the full grid of p-values, apply a multiple-comparison correction, and provide the test statistic or confidence intervals for each gamma.
  4. [Abstract, Sec. 4, Conclusion] The abstract and the conclusion state that experiments were conducted on TIMIT and CHiME datasets, but the experimental section (Section 4) describes the TIMIT and GRID corpora and never mentions CHiME. This is a direct factual inconsistency that obscures the actual scope of the empirical claims. The authors should correct all three places to refer to the same dataset names.
minor comments (5)
  1. [Sec. 2, Fig. 1] The text says 'Kernel Distribution Estimation (KDE)' but the standard term is 'Kernel Density Estimation'; please fix the terminology.
  2. [Sec. 4, Model] The model description says 'a softmax layer with two units that performs the separation task,' while the outputs are described as two 129xM magnitude-spectrum streams. Please clarify how the two-unit softmax produces these output streams, e.g., whether it generates time-frequency masks or directly estimates magnitudes.
  3. [Sec. 3, Eq. 5] The derivation of Eq. (5) silently drops the constant gamma*C and does not state that the expression is a scaled version of Eq. (4). Please add a sentence noting that the constant is irrelevant for optimization and that the scaling is by the fixed gamma.
  4. [Sec. 1 and Sec. 5] There are minor language issues, such as 'which referred to as label permutation ambiguity' in the conclusion and 'we first drive the log-likelihood expression' in Section 3; these should be corrected to 'which is referred to' and 'derive'.
  5. [Sec. 4.1] The sentence 'For each gamma, five experiments are performed (total of 105 experiments for both datasets)' implies 105/5 = 21 gamma values across both datasets, but the visible figure shows fewer discrete gamma values per dataset; please clarify the exact list of gamma values tested.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Prob-PIT's soft-minimum objective is derived from a generative model, and the reported gains are empirical.

full rationale

The derivation chain is self-contained. Equations (1)-(4) define a Gaussian observation model with a uniform prior over output permutations; taking the log-likelihood and applying the log-sum-exp stabilization yields a soft-minimum over permutation costs, with PIT recovered as gamma approaches 0. The paper's central claim is an experimental comparison on TIMIT and GRID, with gamma varied as a hyperparameter. Even if the gamma-selection protocol is not fully specified or the gamma-scaling of Eq. (5) changes the learning-rate scheduler's behavior across conditions, those are experimental-validity concerns, not circular reasoning. No fitted parameter is renamed as a prediction, no load-bearing conclusion rests solely on the authors' prior work, and the claimed improvement is not encoded in the model assumptions. The appropriate finding is no significant circularity.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central derivation rests on two modeling choices: the Gaussian error distribution (which determines the soft-minimum form) and the uniform prior over permutations. The smoothing parameter gamma is a free hyperparameter tuned on validation data. No new physical entities are introduced.

free parameters (1)
  • gamma = 32 (TIMIT), 45 (GRID)
    Smoothing factor in the soft-minimum; tuned by grid search on validation set. Controls the trade-off between the minimum-cost permutation and all permutations.
assumptions (3)
  • domain assumption Estimation error epsilon is i.i.d. Gaussian with variance sigma squared (equivalently gamma = 2 sigma squared).
    Introduced in Sec. 3, Eq. (1)-(4). Justifies the log-sum-exp soft-minimum; if the error distribution differs, the derivation is not exact.
  • domain assumption All S! permutations are equally likely (uniform prior on Z).
    Sec. 3: 'we assume all these S! forms are possible and all have the same probability of 1/S!'. This is a modeling choice; if the prior is non-uniform, the likelihood changes.
  • domain assumption STFT magnitude spectra with phase borrowed from the mixture is a sufficient representation.
    Sec. 2: 'phase information is borrowed from the mixed signal', standard in speech separation but an approximation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Probabilistic Permutation Invariant Training for Speech Separation." pith.science (2026). https://pith.science/paper/FUT5DYSX

@misc{pith2026190801768,
  author       = {Pith},
  title        = {Pith review of: Probabilistic Permutation Invariant Training for Speech Separation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FUT5DYSX}},
  note         = {Machine review of arXiv:1908.01768}
}
read the original abstract

Single-microphone, speaker-independent speech separation is normally performed through two steps: (i) separating the specific speech sources, and (ii) determining the best output-label assignment to find the separation error. The second step is the main obstacle in training neural networks for speech separation. Recently proposed Permutation Invariant Training (PIT) addresses this problem by determining the output-label assignment which minimizes the separation error. In this study, we show that a major drawback of this technique is the overconfident choice of the output-label assignment, especially in the initial steps of training when the network generates unreliable outputs. To solve this problem, we propose Probabilistic PIT (Prob-PIT) which considers the output-label permutation as a discrete latent random variable with a uniform prior distribution. Prob-PIT defines a log-likelihood function based on the prior distributions and the separation errors of all permutations; it trains the speech separation networks by maximizing the log-likelihood function. Prob-PIT can be easily implemented by replacing the minimum function of PIT with a soft-minimum function. We evaluate our approach for speech separation on both TIMIT and CHiME datasets. The results show that the proposed method significantly outperforms PIT in terms of Signal to Distortion Ratio and Signal to Interference Ratio.

Figures

Figures reproduced from arXiv: 1908.01768 by the authors.

Figure 1
Figure 1. Preliminary experiment on the TIMIT dataset for speech separation of a dual-speaker mixed signal using PIT. Cost1 and cost2 are the separation errors for the two possible output-target assignments. Cost1 and cost2 are highly likely to have close values in the first epoch. 3. Probabilistic Permutation Invariant Training (Prob-PIT) This section explains the details of the proposed Prob-PIT modeling technique. Assume X… view at source ↗
Figure 2
Figure 2. Speech separation block diagram. G(.) is a LSTM￾based model and Z(.) is a one-to-one permutation function. log-likelihood expression of the model. According to the model explained in Eq. (1) and considering the Gaussian distribution of , the probability of a clean speech, X, given a mixture sig￾nal, Y , and a specific permutation Z can be written as: P(X|Z, Y ) = N (Z(G(Y, θ)), σ 2 I), (2) where, I is the identity … view at source ↗
Figure 3
Figure 3. Speech separation performance on the TIMIT and GRID datasets. For each γ, five experiments are performed (total of 105 experiments for both datasets). Each violin-shaped object represents the boxplot and the kernel distribution estimation of those five experiments. The blue dashed line, which is the mean of the results obtained for γ = 0, shows the conventional PIT baseline. The red circles on the red solid line are… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 38 canonical work pages

  1. [31]

    Single-channel multi-speaker separation using deep clustering,

    Y . Isik, J. L. Roux, Z. Chen, S. Watanabe, and J. R. Hershey, “Single-channel multi-speaker separation using deep clustering,” arXiv preprint arXiv:1607.02173, 2016

    arXiv 2016

  2. [1]

    Probabilistic Permutation Invariant Training for Speech Separation

    Introduction Humans are equipped with effective abilities to efficiently focus on a particular sound received through their auditory system [1]. In the cocktail party scenario, humans are able to isolate a tar- get speech signal from a mixture of conversations with little effort [2, 3]. However, speech separation is still a challenging task for machines [4...

    work page Pith review arXiv 1908

  3. [2]

    Problem setup and preliminary experiment In single-channel speech separation, we assume that the speech signals have been linearly mixed:y[n] =∑S s=1xs[n], whereS is the number of sources; the goal is to extract all speech signals (i.e.,{xs}S s=1) from the mixed-signal (i.e., y). To do so, sig- nals are normally transferred to the frequency-domain using t...

    work page

  4. [3]

    Probabilistic Permutation Invariant Training (Prob-PIT) This section explains the details of the proposed Prob-PIT modeling technique. Assume Xs is a high-dimensional vec- tor containing the magnitude spectra of the s-th source; X = [X1,...,X S] is a sequence of all Xs vectors; and Y contains the magnitude spectra of the mixed signal. In a model-based spe...

    work page

  5. [4]

    The GRID is a multi-speaker, sentence corpus [32], which has been used in monaural speech separation and recogni- tion challenge [33]

    Experiments, results and discussion Dataset – To evaluate the effectiveness of Prob-PIT, several experiments are conducted on the TIMIT and GRID datasets. The GRID is a multi-speaker, sentence corpus [32], which has been used in monaural speech separation and recogni- tion challenge [33]. Additionally, this corpus has been widely used for assessing the pe...

    work page

  6. [5]

    A long-lasting problem in speech separation task is finding the correct label for each separated speech signal, which referred to as label permutation ambiguity

    Conclusion In this study, we proposed the probabilistic PIT to address the single-channel, speaker-independent speech separation. A long-lasting problem in speech separation task is finding the correct label for each separated speech signal, which referred to as label permutation ambiguity . Recently proposed PIT solves this challenge by training a neural ...

    work page

  7. [6]

    The perception of speech under adverse conditions,

    P. Assmann and Q. Summerfield, “The perception of speech under adverse conditions,” in Speech processing in the auditory system. Springer, 2004, pp. 231–308

    work page 2004

  8. [7]

    A. S. Bregman, Auditory scene analysis: The perceptual organi- zation of sound. MIT press, 1994

    work page 1994

Show all 44 references
  1. [8]

    Auditory grouping, i in hearing. hand- book of perception and cognition, bcj moore,

    C. Darwin and R. Carlyon, “Auditory grouping, i in hearing. hand- book of perception and cognition, bcj moore,” 1995

    1995

  2. [9]

    Divenyi, Speech separation by humans and machines

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

    2004

  3. [10]

    An algorithm to increase intelligibility for hearing-impaired lis- teners in the presence of a competing talker,

    E. W. Healy, M. Delfarah, J. L. Vasko, B. L. Carter, and D. Wang, “An algorithm to increase intelligibility for hearing-impaired lis- teners in the presence of a competing talker,” Journal of Acousti- cal Society of America, vol. 141, no. 6, pp. 4230–4239, 2017

    2017

  4. [11]

    Eeg-informed attended speaker extraction from recorded speech mixtures with application in neuro-steered hearing prostheses,

    S. Van Eyndhoven, T. Francart, and A. Bertrand, “Eeg-informed attended speaker extraction from recorded speech mixtures with application in neuro-steered hearing prostheses,” IEEE Trans. on Biomedical Engineering, vol. 64, no. 5, pp. 1045–1056, 2017

    2017

  5. [12]

    Deep recurrent networks for separation and recognition of single- channel speech in nonstationary background audio,

    H. Erdogan, J. R. Hershey, S. Watanabe, and J. Le Roux, “Deep recurrent networks for separation and recognition of single- channel speech in nonstationary background audio,” in New Era for Robust Speech Recognition. Springer, 2017, pp. 165–186

    2017

  6. [13]

    Single-channel mul- titalker speech recognition,

    S. J. Rennie, J. R. Hershey, and P. A. Olsen, “Single-channel mul- titalker speech recognition,” IEEE Signal Processing Magazine , vol. 27, no. 6, pp. 66–80, 2010

    2010

  7. [14]

    Deep neu- ral networks for single-channel multi-talker speech recognition,

    C. Weng, D. Yu, M. L. Seltzer, and J. Droppo, “Deep neu- ral networks for single-channel multi-talker speech recognition,” IEEE/ACM Transactions on Audio, Speech and Language Pro- cessing (TASLP), vol. 23, no. 10, pp. 1670–1679, 2015

    2015

  8. [15]

    Multi-speaker conversations, cross-talk, and diarization for speaker recognition,

    G. Sell and A. McCree, “Multi-speaker conversations, cross-talk, and diarization for speaker recognition,” in 2017 IEEE Interna- tional Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2017, pp. 5425–5429

    2017

  9. [16]

    All-neural online source sep- aration, counting, and diarization for meeting analysis,

    T. von Neumann, K. Kinoshita, M. Delcroix, S. Araki, T. Nakatani, and R. Haeb-Umbach, “All-neural online source sep- aration, counting, and diarization for meeting analysis,” arXiv preprint arXiv:1902.07881, 2019

    1902 arXiv

  10. [17]

    Jointly aligning and predicting continuous emotion annotations,

    S. Khorram, M. McInnis, and E. M. Provost, “Jointly aligning and predicting continuous emotion annotations,” IEEE Transactions on Affective Computing, 2019

    2019

  11. [18]

    Progressive neural networks for transfer learning in emotion recognition,

    J. Gideon, S. Khorram, Z. Aldeneh, D. Dimitriadis, and E. M. Provost, “Progressive neural networks for transfer learning in emotion recognition,”arXiv preprint arXiv:1706.03256, 2017

    2017 arXiv

  12. [19]

    A robust text depen- dent speaker identification using neural responses from the model of the auditory system,

    M. I. Khalil, N. Mamun, and K. Akter, “A robust text depen- dent speaker identification using neural responses from the model of the auditory system,” in International Conference on ECCE . IEEE, 2019, pp. 1–4

    2019

  13. [20]

    Speaker recognition by machines and humans: A tutorial review,

    J. H. Hansen and T. Hasan, “Speaker recognition by machines and humans: A tutorial review,” IEEE Signal processing magazine , vol. 32, no. 6, pp. 74–99, 2015

    2015

  14. [21]

    Computational auditory scene analysis: Principles, algorithms, application

    D. Wang and G.Brown, “Computational auditory scene analysis: Principles, algorithms, application.”Wiley-IEEEPress, 2006

    2006

  15. [22]

    Independent component analysis, a new concept?

    P. Comon, “Independent component analysis, a new concept?” Signal processing, vol. 36, no. 3, pp. 287–314, 1994

    1994

  16. [23]

    One microphone source separation,

    S. T. Roweis, “One microphone source separation,” in Advances in neural information processing systems, 2001, pp. 793–799

    2001

  17. [24]

    Convolutive speech bases and their applica- tion to supervised speech separation,

    P. Smaragdis et al., “Convolutive speech bases and their applica- tion to supervised speech separation,” IEEE Transactions on au- dio speech and language processing, vol. 15, no. 1, p. 1, 2007

    2007

  18. [25]

    Supervised speech enhancement using online group-sparse convolutive nmf,

    M. Yousefi and M. H. Savoji, “Supervised speech enhancement using online group-sparse convolutive nmf,” inInternational Sym- posium on Telecommunications (IST). IEEE, 2016, pp. 494–499

    2016

  19. [26]

    On training targets for supervised speech separation,

    Y . Wang, A. Narayanan, and D. Wang, “On training targets for supervised speech separation,” IEEE/ACM Transactions on Au- dio, Speech and Language Processing (TASLP) , vol. 22, no. 12, pp. 1849–1858, 2014

    2014

  20. [27]

    Joint optimization of masks and deep recurrent neural networks for monaural source separation,

    P.-S. Huang, M. Kim, M. Hasegawa-Johnson, and P. Smaragdis, “Joint optimization of masks and deep recurrent neural networks for monaural source separation,” IEEE/ACM Trans. on Audio, Speech, and Lang. Proc., vol. 23, no. 12, pp. 2136–2147, 2015

    2015

  21. [28]

    A deep ensemble learning method for monaural speech separation,

    X.-L. Zhang and D. Wang, “A deep ensemble learning method for monaural speech separation,”IEEE/ACM Transactions on Au- dio, Speech and Language Processing (TASLP), vol. 24, no. 5, pp. 967–977, 2016

    2016

  22. [29]

    Convolutional neu- ral network-based speech enhancement for cochlear implant re- cipients,

    N. Mamun, S. Khorram, and J. H. L. Hansen, “Convolutional neu- ral network-based speech enhancement for cochlear implant re- cipients,” inProc. Interspeech, 2019

    2019

  23. [30]

    Deep clus- tering: Discriminative embeddings for segmentation and separa- tion,

    J. R. Hershey, Z. Chen, J. Le Roux, and S. Watanabe, “Deep clus- tering: Discriminative embeddings for segmentation and separa- tion,” inAcoustics, Speech and Signal Processing (ICASSP), 2016 IEEE International Conference on. IEEE, 2016, pp. 31–35

    2016

  24. [32]

    Deep attractor network for single-microphone speaker separation,

    Z. Chen, Y . Luo, and N. Mesgarani, “Deep attractor network for single-microphone speaker separation,” in Acoustics, Speech and Signal Processing (ICASSP), 2017 IEEE International Confer- ence on. IEEE, 2017, pp. 246–250

    2017

  25. [33]

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

    D. Yu, M. Kolbæk, Z.-H. Tan, and J. Jensen, “Permutation invari- ant training of deep models for speaker-independent multi-talker speech separation,” in Acoustics, Speech and Signal Processing (ICASSP), 2017 IEEE International Conference on. IEEE, 2017, pp. 241–245

    2017

  26. [34]

    Multitalker speech separation with utterance- level permutation invariant training of deep recurrent neural net- works,

    M. Kolbæk, D. Yu, Z.-H. Tan, J. Jensen, M. Kolbaek, D. Yu, Z.-H. Tan, and J. Jensen, “Multitalker speech separation with utterance- level permutation invariant training of deep recurrent neural net- works,”IEEE/ACM Transactions on Audio, Speech and Language Processing (TASLP),...

    1901

  27. [35]

    Complex ratio mask- ing for monaural speech separation,

    D. S. Williamson, Y . Wang, and D. Wang, “Complex ratio mask- ing for monaural speech separation,” IEEE/ACM Transactions on Audio, Speech and Language Processing (TASLP), vol. 24, no. 3, pp. 483–492, 2016

    2016

  28. [36]

    Soft-dtw: a differentiable loss func- tion for time-series,

    M. Cuturi and M. Blondel, “Soft-dtw: a differentiable loss func- tion for time-series,” in Proceedings of the 34th International Conference on Machine Learning-Volume 70. JMLR. org, 2017, pp. 894–903

    2017

  29. [37]

    An audio- visual corpus for speech perception and automatic speech recog- nition,

    M. Cooke, J. Barker, S. Cunningham, and X. Shao, “An audio- visual corpus for speech perception and automatic speech recog- nition,” The Journal of the Acoustical Society of America , vol. 120, no. 5, pp. 2421–2424, 2006

    2006

  30. [38]

    Monaural speech separation and recognition challenge,

    M. Cooke, J. R. Hershey, and S. J. Rennie, “Monaural speech separation and recognition challenge,” Computer Speech & Lan- guage, vol. 24, no. 1, pp. 1–15, 2010

    2010

  31. [39]

    Teager–kaiser energy operators for overlapped speech detection,

    N. Shokouhi and J. H. Hansen, “Teager–kaiser energy operators for overlapped speech detection,”IEEE/ACM Transactions on Au- dio, Speech, and Language Processing, vol. 25, no. 5, pp. 1035– 1047, 2017

    2017

  32. [40]

    Assessing speaker en- gagement in 2-person debates: Overlap detection in united states presidential debates,

    M. Yousefi, N. Shokouhi, and J. Hansen, “Assessing speaker en- gagement in 2-person debates: Overlap detection in united states presidential debates,” inProc. Interspeech, 2018, pp. 2117–2121

    2018

  33. [41]

    Speech separation based on signal-noise-dependent deep neural networks for robust speech recognition,

    Y .-H. Tu, J. Du, L.-R. Dai, and C.-H. Lee, “Speech separation based on signal-noise-dependent deep neural networks for robust speech recognition,” in 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2015, pp. 61–65

    2015

  34. [42]

    Performance measure- ment in blind audio source separation,

    E. Vincent, R. Gribonval, and C. F ´evotte, “Performance measure- ment in blind audio source separation,” IEEE trans. on audio, speech, and lang. processing, vol. 14, no. 4, pp. 1462–1469, 2006

    2006

  35. [43]

    Modeling perceptual similarity of audio signals for blind source separation evaluation,

    B. Fox, A. Sabin, B. Pardo, and A. Zopf, “Modeling perceptual similarity of audio signals for blind source separation evaluation,” in International Conference on Independent Component Analysis and Signal Separation. Springer, 2007, pp. 454–461

    2007

  36. [44]

    Long short-term memory for speaker gen- eralization in supervised speech separation,

    J. Chen and D. Wang, “Long short-term memory for speaker gen- eralization in supervised speech separation,”Journal of the Acous- tical Society of America, vol. 141, no. 6, pp. 4705–4714, 2017

    2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.