Pith. sign in

REVIEW 3 major objections 5 minor 37 references

A Dual-Staged Context Aggregation Method Towards Efficient End-To-End Speech Enhancement

T0 review · 3 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 This paper proposes DCCRN, a hybrid of a dilated densely connected convolutional network and a compact gated recurrent unit, and argues that this two-stage context aggregation denoises raw speech waveforms well enough to beat…

desk verdict DCCRN is a competent low-cost hybrid for end-to-end enhancement, but the paper's causal claim about the dual-staged design is not actually isolated from the shortcut and staged training. read the letter →

arxiv 1908.06468 v4 pith:XVKE55BA submitted 2019-08-18 cs.SD cs.LGeess.AS

Kai Zhen , Mi Suk Lee , Minje Kim This is my paper · ORCID
classification cs.SDcs.LGeess.AS
keywords end-to-endspeechenhancementcontextaggregationdenseconnectivitydilatedconvolutiongatedrecurrentunitresiduallearningtime-domainprocessingDCCRN
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

DCCRN is an end-to-end speech-enhancement network that works directly on raw time-domain waveforms instead of spectrograms or masks. The paper argues that two complementary context-aggregation mechanisms—a dilated DenseNet that enlarges the receptive field and a compact many-to-one GRU that pools temporal context—can be combined so that the recurrent part only needs to refine an already denoised signal. On the paper's test mixtures with trained noise types and untrained speakers, DCCRN outperforms both convolutional baselines across all reported metrics, with an average STOI improvement of 0.23 and a PESQ improvement of 1.38 over unprocessed speech. The model has 1.38 million parameters, and on untrained noises it remains competitive with a larger Wave-U-Net-style model on intelligibility while falling behind on perceptual quality.

What carries the argument

The carrying object is DCCRN's two-stage pipeline: a dilated DenseNet followed by a compact GRU, joined by an additive identical shortcut. In each dense block, every convolutional layer takes the channel-wise concatenation of all earlier feature maps in the block, so information flows forward without extra parameters; the middle layer of each block uses a dilation rate of 1, 2, 4, or 8 to widen the receptive field. After the four blocks, the frame is split into M sub-frames and fed sequentially to two GRU layers that emit only the last state, producing the denoised output. The identical shortcut adds the final DenseNet sub-frame to the GRU output, turning the recurrent stage into a residual refiner. The training scheme—CNN pretraining, locked-CNN GRU pretraining, then joint fine-tuning with a small learning rate—plus an objective that combines time-domain mean squared error with a mel-spectrogram regularizer makes the heterogeneous model trainable.

What would settle it

Remove the cross-component shortcut from DCCRN and retrain on the identical data and schedule; also train DenseNet+GRU with the same staged schedule and the shortcut while disabling dilation. If either variant matches or beats the reported SDR, STOI, and PESQ numbers, the dual-stage attribution fails.

Watch

Extended reading notes

Core claim

The central claim is that temporal context can be gathered at two successive levels within one small network. A four-block densely connected 1D-CNN first denoises an N-sample window using dilation rates 1, 2, 4, and 8 in the middle layer of each block, so that the receptive field grows without a huge kernel. The output frame is then reshaped into M short sub-frames and read by two GRU layers in a many-to-one fashion, making the final output a function of the whole sequence. An additive cross-component shortcut copies the last DenseNet sub-frame onto the GRU output, so the recurrent stage needs to supply only a residual correction. With this design and a schedule that pretrains the CNN, then the GRU, then fine-tunes both, the paper reports better signal-to-distortion ratio, STOI, and PESQ than a dilated DenseNet alone or a DenseNet followed by GRU without dilation on unseen speakers at -5, 0, and +5 dB SNR. On unseen noises it reports intelligibility close to a much larger Wave-U-Net while remaining around one seventh of its parameter count.

Load-bearing premise

The paper's performance claim stands or falls on the assumption that the improvement over baselines is caused by the dual-stage aggregation and the cross-component shortcut, but no ablation isolates those parts, so the extra GRU capacity or the staged pretraining could be the actual cause.

Editorial extensions

If this is right

  • The full 1.38-million-parameter model reports higher SDR, STOI, and PESQ than both convolutional baselines on unseen speakers at -5, 0, and +5 dB, so dual-stage aggregation appears compatible with affordable end-to-end enhancement.
  • Because the network emits only the last M-sample sub-frame and uses no future samples, the latency is one sub-frame—16 ms at 16 kHz—suggesting the design is usable in near-real-time settings.
  • Enlarging the frame to 4,096 samples does not improve results, indicating that beyond some point the recurrent stage's optimization difficulty outweighs the benefit of longer context.
  • On untrained noise types, the model's intelligibility is close to or better than Wave-U-Net at some SNRs despite the parameter gap, while its perceptual quality trails; the paper frames this as the expected complexity trade-off.

Reading between the lines

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

  • A decisive follow-up experiment would give the DenseNet+GRU baseline the same staged schedule and the same shortcut while disabling dilation: if it matches DCCRN, the improvement comes from training and gating, not from dilated context aggregation.
  • Because the shortcut makes the GRU a residual corrector, one could shrink the GRU hidden size or prune it after CNN pretraining; the parameter budget could likely go lower than 1.38 million, though the paper does not test this.
  • The two-stage pattern could transfer to other waveform-to-waveform tasks—speech coding, source separation, dereverberation—by substituting the perceptual regularizer, since the mel-spectrogram term already provides the needed training signal.
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

3 major / 5 minor

Summary. The paper proposes DCCRN, a hybrid time-domain end-to-end speech enhancement network that cascades a dilated DenseNet with a two-layer GRU. A cross-component additive shortcut carries the DenseNet output to the GRU output, and a staged training procedure pretrains the CNN and RNN components before joint fine-tuning. Experiments on TIMIT mixtures with untrained speakers at -5, 0, and +5 dB SNR compare DCCRN with Dilated DenseNet and DenseNet+GRU baselines using SDR, SIR, SAR, STOI, and PESQ; a second experiment on untrained noise types compares DCCRN with Wave-U-Net reported in prior work. The paper reports consistent improvements of DCCRN over both convolutional baselines, with 1.38 million parameters.

Significance. If the empirical results are reproducible, DCCRN is a useful contribution: a compact end-to-end model that outperforms convolutional and convolutional-recurrent baselines on untrained speakers while remaining small enough for resource-constrained use, and with plausible but weaker generalization to untrained noises than a much larger Wave-U-Net. The use of public corpora and a provided demo page is a strength. However, the paper's mechanistic claim that dual-staged aggregation and the shortcut drive the gains is not established by the reported experiments, and the cross-system generalization comparison is not fully matched. These issues are fixable with additional experiments and reporting, so the contribution is potentially acceptable after major revision.

major comments (3)
  1. [§3.2, Table 2, §2.5] The headline claim that dual-staged context aggregation and the cross-component identical shortcut produce the observed gains is not supported, because DCCRN differs from Dilated DenseNet in three ways (the GRU component, the additive shortcut, and the three-phase training scheme) and from DenseNet+GRU in at least two ways (dilated middle layers and the additive shortcut), while the training scheme in Section 2.5 is described only for DCCRN. I request ablations that isolate each factor: DCCRN without the shortcut, DCCRN trained end-to-end without the staged schedule, and DenseNet+GRU trained with the same staged schedule (or DCCRN without dilation). Without these, the abstract's causal phrasing is unjustified.
  2. [§3.2, Table 2] No measure of variability is reported. The STOI and PESQ differences between DCCRN and DenseNet+GRU are small (e.g., STOI 0.92 vs 0.90 at -5 dB and 0.96 vs 0.95 at +5 dB; PESQ 2.14 vs 1.96 at -5 dB), and the test set uses a random selection of unseen utterances. Please report standard deviations or confidence intervals across random training/test draws, or paired significance tests, so that 'consistently outperforms' is supported beyond a single run.
  3. [§3.4, Fig. 2] The comparison with Wave-U-Net is imported from [29] under a different evaluation protocol: [29] reports narrowband PESQ, whereas DCCRN is evaluated in wideband mode (Section 3.3), and the SNR levels (-3, +3, +6 dB) and the noise/training conditions are not stated to be identical. The paper should either re-run Wave-U-Net under the same protocol or present this only as a qualitative complexity trade-off, not as a head-to-head quality comparison.
minor comments (5)
  1. [Abstract] The average STOI improvement of 0.23 and PESQ of 1.38 are stated immediately after 'consistently outperforms competing convolutional baselines', which invites reading them as improvements over the baselines; Section 3.3 clarifies that they are improvements over unprocessed mixtures. Reword the abstract.
  2. [§2.1, Eq. (2)] The convolution definition omits the summation over input channels; since H^(l) is later defined as R^{K x (l - l_b + 1)D}, the equation should sum over the input-channel index as well.
  3. [Algorithm 1, line 2] 'The last M/N samples' should read 'the last N/M samples'; line 4 uses the correct ratio.
  4. [§3.1, reference [24]] The five non-stationary noise types (birds, cicadas, computer keyboard, machine guns, motorcycles) are cited to [24], which is the BSS Eval paper; a noise-corpus reference is needed.
  5. [§3.3] The statement that the average SDR improvement over the DenseNet baseline is 17.3% is inconsistent with the values in Table 2, which imply an average relative improvement of about 11.5% (or 1.78 dB absolute); please recalculate or clarify the reference value.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's claims are empirical comparisons against external baselines; Eq. (8) is a training loss, not a derived prediction.

full rationale

The paper makes no analytic derivation that could reduce to its inputs. DCCRN is an empirical architecture; the objective in Eq. (8) is MSE plus a mel-spectrogram regularizer and is used as a training loss, not as a quantity derived from the evaluation metrics. The only self-citation, [22], supports this auxiliary regularizer and is not load-bearing for the claimed architecture-level contribution. Table 2 and Fig. 2 compare DCCRN with external baselines on TIMIT using independent BSS Eval, STOI, and PESQ metrics, and no parameter is fitted to those benchmarks to produce the reported STOI/PESQ improvements. The abstract's average improvements compare DCCRN output with unprocessed mixtures on a held-out test set, so the comparison is external. The lack of an ablation separating the shortcut, dilation, and staged training is a genuine experimental-validity concern, but it is a confound in the empirical comparison rather than a circular step: no equation or fitted parameter is equivalent by construction to the claimed result.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard supervised learning assumptions plus several design choices. The free parameters are architecture and training hyperparameters chosen by hand; no physical or fitted constants are involved. The axioms are domain assumptions about dataset representativeness, metric validity, and fair comparison. No invented entities are introduced.

free parameters (7)
  • DenseNet growth rate D = 32
    Sets the number of channels added per dense layer; hand-chosen capacity knob.
  • Dilation rates = 1, 2, 4, 8
    Middle-layer dilation rate of the four dense blocks; hand-chosen receptive field schedule.
  • Kernel sizes = 5 and 55
    Small kernels in dense layers and a large kernel in the middle layer; hand-chosen to enlarge the receptive field.
  • Frame size N and sub-frame count M = N=1024, M=4
    Context window length and GRU sequence length; hand-chosen. A larger N=4096 variant was tested but did not improve.
  • GRU hidden units = 32 and 256
    The first GRU layer is compact; the second matches output dimensionality N/M; hand-chosen.
  • Mel regularizer weight lambda = 1/60
    Weights the mel-spectrogram MSE term in Eq. (8); chosen empirically.
  • Learning rates = 1e-4, 5e-6, 5e-7
    Three-stage training schedule for CNN, RNN, and finetune phases; chosen by hand.
assumptions (4)
  • domain assumption MSE plus mel-spectrogram MSE in Eq. (8) is a valid proxy for the evaluated STOI and PESQ objectives.
    The model is trained to minimize time-domain and mel-spectral MSE, while success is measured by STOI/PESQ; the correlation is borrowed from prior work and not independently verified here.
  • domain assumption The TIMIT 100-utterance test subset and five selected noise types adequately represent untrained-speaker conditions.
    Section 3.1 randomly selects one test split; no repeated splits or confidence intervals are reported, so small-sample variation could affect the ranked comparisons.
  • domain assumption The Wave-U-Net numbers from [29] are compatible with DCCRN's evaluation protocol for the complexity comparison in Fig. 2.
    Section 3.4 states that Wave-U-Net was reported in narrowband mode, while DCCRN uses wide-band PESQ in Table 2, so the two sets may not be directly comparable.
  • domain assumption Adam training to 'best effort' gives all compared models equally close to their optima.
    Section 3.2 states 'all models are trained to our best effort', but tuning effort is not quantified; baseline performance could depend on search effort.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Dual-Staged Context Aggregation Method Towards Efficient End-To-End Speech Enhancement." pith.science (2026). https://pith.science/paper/XVKE55BA

@misc{pith2026190806468,
  author       = {Pith},
  title        = {Pith review of: A Dual-Staged Context Aggregation Method Towards Efficient End-To-End Speech Enhancement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XVKE55BA}},
  note         = {Machine review of arXiv:1908.06468}
}
read the original abstract

In speech enhancement, an end-to-end deep neural network converts a noisy speech signal to a clean speech directly in time domain without time-frequency transformation or mask estimation. However, aggregating contextual information from a high-resolution time domain signal with an affordable model complexity still remains challenging. In this paper, we propose a densely connected convolutional and recurrent network (DCCRN), a hybrid architecture, to enable dual-staged temporal context aggregation. With the dense connectivity and cross-component identical shortcut, DCCRN consistently outperforms competing convolutional baselines with an average STOI improvement of 0.23 and PESQ of 1.38 at three SNR levels. The proposed method is computationally efficient with only 1.38 million parameters. The generalizability performance on the unseen noise types is still decent considering its low complexity, although it is relatively weaker comparing to Wave-U-Net with 7.25 times more parameters.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 30 canonical work pages

  1. [29]

    Performance mea- surement in blind audio source separation,

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

    work page 2006

  2. [1]

    A Dual-Staged Context Aggregation Method Towards Efficient End-To-End Speech Enhancement

    INTRODUCTION Monaural speech enhancement can be described as a process to ex- tract the target speech signal by suppressing the background interfer- ence in the speech mixture in the single-microphone setting. There have been various classic methods, such as spectral subtraction [1], Wiener-filtering [2] and non-negative matrix factorization [3], to re- mo...

    work page Pith review arXiv 2017

  3. [2]

    MODEL DESCRIPTION 2.1. Context aggregation with dilated DenseNet Residual learning has become a critical technique to tackle the gradi- ent vanishing issue when tuning a deep convolutional neural network (CNN), such that the deep CNN can achieve better performance but with a lower model complexity. ResNet illustrates a classic way to enable residual learn...

    work page

  4. [3]

    many-to-one

    This forms our fully convolutional DenseNet baseline. In all the convolutional layers, we use leaky ReLU as the activation. 2.2. Context aggregation with gated recurrent network DCCRN further employs RNN layers following the dilated DenseNet component (Figure 1). Among LSTM and GRU, two most well-known RNN variations, DCCRN chooses GRU for its re- duced c...

    work page

  5. [4]

    Experimental setup In this paper, the experiment runs on TIMIT corpus [23]

    EXPERIMENTS 3.1. Experimental setup In this paper, the experiment runs on TIMIT corpus [23]. We con- sider two experimental settings. For the model training, we ran- domly select 1000 utterances from TIMIT training subset. 5 types of non-stationary noise (birds, cicadas, computer keyboard, machine guns and motorcycles) from [24] are used to create mixture...

    work page

  6. [5]

    DCCRN firstly suppresses the noise in time domain with dilated DenseNet, followed by a GRU component to further leverage the temporal context in a many-to-one manner

    CONCLUSION The paper introduces DCCRN, a hybrid residual network, to aggre- gate temporal context in dual levels for efficient end-to-end speech enhancement. DCCRN firstly suppresses the noise in time domain with dilated DenseNet, followed by a GRU component to further leverage the temporal context in a many-to-one manner. To tune the model with heterogenei...

    work page

  7. [6]

    Suppression of acoustic noise in speech using spec- tral subtraction,

    S. F. Boll, “Suppression of acoustic noise in speech using spec- tral subtraction,” vol. 27, pp. 113–120, 1979

    work page 1979

  8. [7]

    Brown, P Hwang, et al., Introduction to random signals and applied Kalman filtering, vol

    R.G. Brown, P Hwang, et al., Introduction to random signals and applied Kalman filtering, vol. 3, Wiley New York, 1992

    work page 1992

Show all 37 references
  1. [8]

    Single-channel speech separation using sparse non-negative matrix factorization,

    M. Schmidt and R. Olsson, “Single-channel speech separation using sparse non-negative matrix factorization,” inNinth Inter- national Conference on Spoken Language Processing, 2006

    2006

  2. [9]

    Complex ratio masking for monaural speech separation,

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

    2016

  3. [10]

    Incremental binarization on recurrent neural networks for single-channel source sepa- ration,

    S. Kim, M. Maity, and M. Kim, “Incremental binarization on recurrent neural networks for single-channel source sepa- ration,” in 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2019

    2019

  4. [11]

    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 pro- cessing, vol. 27, no. 8, pp. 1256–1266, 2019

    2019

  5. [12]

    A regression approach to speech enhancement based on deep neural net- works,

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

    2014

  6. [13]

    Ideal ratio mask estimation us- ing deep neural networks for robust speech recognition,

    A. Narayanan and D. Wang, “Ideal ratio mask estimation us- ing deep neural networks for robust speech recognition,” in 2013 IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2013, pp. 7092–7096

    2013

  7. [14]

    Segan: Speech en- hancement generative adversarial network,

    S. Pascual, A. Bonafonte, and J. Serr `a, “Segan: Speech en- hancement generative adversarial network,” arXiv preprint arXiv:1703.09452, 2017

    2017 arXiv

  8. [15]

    On ideal binary mask as the computational goal of auditory scene analysis,

    D. Wang, “On ideal binary mask as the computational goal of auditory scene analysis,” in Speech separation by humans and machines, pp. 181–197. Springer, 2005

    2005

  9. [16]

    Multi-scale context aggregation by di- lated convolutions,

    F. Yu and V . Koltun, “Multi-scale context aggregation by di- lated convolutions,” arXiv preprint arXiv:1511.07122, 2015

    2015 arXiv

  10. [17]

    Gated residual networks with dilated convolutions for supervised speech separation,

    K. Tan, J. Chen, and D. Wang, “Gated residual networks with dilated convolutions for supervised speech separation,” in Proc. ICASSP, 2018

    2018

  11. [18]

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

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

    2017

  12. [19]

    Wavenet: A generative model for raw au- dio,

    A. v. d. Oord, S. Dieleman, H. Zen, K. Simonyan, O. Vinyals, A. Graves, et al., “Wavenet: A generative model for raw au- dio,” arXiv preprint arXiv:1609.03499, 2016

    2016 arXiv

  13. [20]

    A convolutional recurrent neural net- work for real-time speech enhancement,

    K. Tan and D. Wang, “A convolutional recurrent neural net- work for real-time speech enhancement,” in Proceedings of Interspeech, 2018, pp. 3229–3233

    2018

  14. [21]

    Densely connected network with time-frequency dilated convolution for speech enhancement,

    Y . Li, X. Li, Y . Dong, M. Li, S. Xu, and S. Xiong, “Densely connected network with time-frequency dilated convolution for speech enhancement,” in ICASSP 2019-2019 IEEE Interna- tional Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2019, pp. 6860–6864

    2019

  15. [22]

    Mmdenselstm: An efficient combination of convolutional and recurrent neural networks for audio source separation,

    N. Takahashi, N. Goswami, and Y . Mitsufuji, “Mmdenselstm: An efficient combination of convolutional and recurrent neural networks for audio source separation,” in 2018 16th Interna- tional Workshop on Acoustic Signal Enhancement (IWAENC). IEEE, 2018, pp. 106–110

    2018

  16. [23]

    Densely connected convolutional networks,

    G. Huang, Z. Liu, Laurens Van D. M., and K. Q Weinberger, “Densely connected convolutional networks,” in Proceedings of the IEEE conference on computer vision and pattern recog- nition, 2017, pp. 4700–4708

    2017

  17. [24]

    Empiri- cal evaluation of gated recurrent neural networks on sequence modeling,

    J. Chung, C. Gulcehre, K. Cho, and Y . Bengio, “Empiri- cal evaluation of gated recurrent neural networks on sequence modeling,” arXiv preprint arXiv:1412.3555, 2014

    2014 arXiv

  18. [25]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016

    2016

  19. [26]

    End- to-end waveform utterance enhancement for direct evaluation metrics optimization by fully convolutional neural networks,

    S. Fu, T. Wang, Y . Tsao, X. Lu, and H. Kawai, “End- to-end waveform utterance enhancement for direct evaluation metrics optimization by fully convolutional neural networks,” IEEE/ACM Transactions on Audio, Speech and Language Pro- cessing (TASLP), vol. 26, no. 9, pp. 1570–1584, 2018

    2018

  20. [27]

    Cascaded Cross-Module Residual Learning Towards Lightweight End- to-End Speech Coding,

    K. Zhen, J. Sung, M. S. Lee, S. Beack, and M. Kim, “Cascaded Cross-Module Residual Learning Towards Lightweight End- to-End Speech Coding,” in Proceedings of Interspeech 2019

    2019

  21. [28]

    Darpa timit acoustic-phonetic continous speech cor- pus cd-rom. nist speech disc 1-1.1,

    J. S Garofolo, L. F Lamel, W. M Fisher, J. G Fiscus, and D. S Pallett, “Darpa timit acoustic-phonetic continous speech cor- pus cd-rom. nist speech disc 1-1.1,”NASA STI/Recon technical report n, vol. 93, 1993

    1993

  22. [30]

    Adam: A method for stochastic opti- mization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic opti- mization,” arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  23. [31]

    An algorithm for intelligibility prediction of time–frequency weighted noisy speech,

    C. H Taal, R. C Hendriks, R. Heusdens, and J. Jensen, “An algorithm for intelligibility prediction of time–frequency weighted noisy speech,” IEEE Transactions on Audio, Speech, and Language Processing, vol. 19, no. 7, pp. 2125–2136, 2011

    2011

  24. [32]

    Per- ceptual 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, “Per- ceptual 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 (ICASSP). IEEE

    2001

  25. [33]

    Perceptual evaluation of speech quality (PESQ): an objective method for end-to-end speech quality assessment of narrow-band telephone networks and speech codecs,

    Recommendation P ITU-T, “Perceptual evaluation of speech quality (PESQ): an objective method for end-to-end speech quality assessment of narrow-band telephone networks and speech codecs,” International Telecommunication Union- Telecommunication Standardisation Sector, 2001

    2001

  26. [34]

    Incorporating sym- bolic sequential modeling for speech enhancement,

    C. Liao, Y . Tsao, X. Lu, and H. Kawai, “Incorporating sym- bolic sequential modeling for speech enhancement,” arXiv preprint arXiv:1904.13142, 2019

    1904 arXiv

  27. [35]

    100 nonspeech environmental sounds,

    G. Hu, “100 nonspeech environmental sounds,” The Ohio State University, 2004

    2004

  28. [36]

    Noisex-92: A database and an experiment to study the effect of additive noise on speech recognition systems,

    A. Varga and J. H. Steeneken, “Noisex-92: A database and an experiment to study the effect of additive noise on speech recognition systems,” Speech communication, vol. 12, no. 3, pp. 247–251, 1993

    1993

  29. [37]

    Wave-U-Net: A multi- scale neural network for end-to-end audio source separation,

    D. Stoller, S. Ewert, and S. Dixon, “Wave-U-Net: A multi- scale neural network for end-to-end audio source separation,” arXiv preprint arXiv:1806.03185, 2018

    2018 arXiv

Pith tools

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