Pith. sign in

REVIEW 3 major objections 8 minor 1 cited by

EchoFree: Towards Ultra Lightweight and Efficient Neural Acoustic Echo Cancellation

T0 review · 3 major / 8 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read EchoFree, a hybrid linear-plus-neural acoustic echo canceller with 278K parameters and 30 MMACs/s, reports AECMOS scores on the ICASSP 2023 blind test that match the much larger DeepVQE-S while beating two low-complexity baselines.

desk verdict A genuinely tiny hybrid AEC that lands in DeepVQE-S territory on the ICASSP 2023 blind test, but the 'surpasses' claim rests on single-run AECMOS margins that could easily be noise. read the letter →

arxiv 2508.06271 v1 pith:7IVUK7SJ submitted 2025-08-08 eess.AS

Xingchen Li , Boyi Kang , Ziqian Wang , Zihan Zhang , Mingshuai Liu , Zhonghua Fu , Lei Xie This is my paper · ORCID
classification eess.AS
keywords acousticechocancellationBark-scalefeaturesneuralpost-filterself-supervisedlearningWavLMlightweightmodelhybridAECAECMOS
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 tries to establish that a well-chosen hybrid architecture—an adaptive linear filter for the bulk of the echo, and a very small neural post-filter working on perceptually compressed Bark-scale spectra—can reach the quality of much larger neural AEC systems. The model, EchoFree, has only 278K parameters and requires 30 MMACs per second of audio, a small fraction of the compute and size of the models it is compared with. On the ICASSP 2023 blind test, it reports AECMOS scores above two lightweight baselines and comparable to DeepVQE-S, the strongest small model in the comparison. The paper attributes the efficiency to Bark-scale feature compression and the quality to a two-stage training scheme that first aligns the post-filter to frozen WavLM embeddings, then finetunes it with a Bark-domain gain loss. If the results hold, resource-constrained devices can run high-quality echo cancellation in real time.

What carries the argument

The central mechanism is the pairing of a compact spectral representation with a two-stage perceptual training objective. The Bark-scale feature extractor multiplies the 257-bin magnitude spectrum by a fixed 257-by-100 mapping matrix, takes log power, and appends first and second time-derivatives of the first six bands, yielding a 112-dimensional input per frame; this compression is what lets the whole post-filter run at 30 MMACs/s. The two-stage loss uses frozen WavLM-Large embeddings (layer-wise MSE) in stage one and then a Bark-scale gain loss (quartic plus quadratic terms with a cross-entropy regularizer) in stage two, so the model first learns coarse speech structure and then fine perce

What would settle it

Bootstrap AECMOS over the 800 blind-test samples using the same Run 1668423760 Stage 0.onnx checkpoint and compute confidence intervals for the EchoFree-minus-DeepVQE-S difference in EchoMOS and DegMOS for each scenario (ST FE, ST NE, DT). If any interval includes zero, the claim that EchoFree 'surpasses' DeepVQE-S is not supported. A second, independent check is to run both systems on the ICASSP 2024 AEC Challenge blind set and see whether EchoFree again lands within a few hundredths of DeepVQE-S.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that a neural post-filter of 278K parameters, operating on 100 Bark-scale log-power bands, suppresses the residual echo left by a partitioned-block frequency-domain adaptive Kalman filter almost as well as DeepVQE-S, a model with roughly three times the parameters and ten times the compute. The post-filter is a depthwise-separable U-Net with a unidirectional GRU bottleneck and subpixel upsampling; it outputs Bark-domain gains, converted back to a linear-frequency magnitude mask via the transpose of the Bark mapping matrix. Training is the distinctive part: stage one minimizes layer-wise mean-squared error between the estimated and ground-truth s

Load-bearing premise

The efficiency and quality comparison depends on AECMOS scores from a single evaluation checkpoint, with no repeated runs or confidence intervals; if AECMOS's sample-level noise is on the order of 0.1, the claimed edge over DeepVQE-S is within the noise.

Editorial extensions

If this is right

  • At 278K params and 30 MMACs/s, a device that can afford a traditional adaptive filter can also afford EchoFree's post-filter, making near-SOTA echo cancellation feasible on phones and hearing aids.
  • The two-stage SSL-plus-gain training recipe shows that a frozen large model can teach a tiny network, with no SSL cost at inference.
  • The gains over ULCNet-AER and Bark-AEC suggest that low-complexity AEC models can be compressed further without losing the quality gap to end-to-end systems.
  • The model's streaming-friendly design, with a causal GRU bottleneck and frame-wise Bark features, supports real-time use rather than offline processing.
  • If replicated, the approach gives a concrete efficiency target for the AEC community: match DeepVQE-S with 30 MMACs/s rather than 315.

Reading between the lines

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

  • One extension the paper leaves implicit: because WavLM is frozen and discarded at inference, the same two-stage distillation idea could be applied with any SSL representation model, possibly yielding further gains as those models improve.
  • The reported quality differences between EchoFree and DeepVQE-S are within typical AECMOS variability, so a direct bootstrap or repeated-checkpoint comparison would clarify whether the 'surpasses' wording is justified; the paper does not provide that analysis.
  • The dynamic RIR/SER simulation pipeline, though not evaluated for generalization, suggests the model's robustness could be tested on later Challenge blind sets or on real device recordings.
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 / 8 minor

Summary. The paper proposes EchoFree, a hybrid acoustic echo cancellation (AEC) system that cascades a partitioned-block frequency-domain adaptive Kalman filter with an ultra-lightweight neural post-filter operating on Bark-scale spectral features. The neural post-filter uses a U-Net-like encoder/decoder with depthwise separable convolutions and a GRU bottleneck, totaling 278K parameters and 30 MMACs/s. Training uses a two-stage strategy: first, an SSL loss based on frozen WavLM-Large embeddings; second, a combination of that SSL loss with a Bark-scale gain loss. The method is evaluated on the ICASSP 2023 AEC Challenge blind test set using the AECMOS metric. The authors claim the model outperforms existing low-complexity AEC models and achieves performance comparable to the much larger DeepVQE-S, establishing a favorable efficiency/performance trade-off.

Significance. If the reported efficiency-performance trade-off is reliable, this is a useful contribution to low-resource AEC: a very small neural post-filter that scores competitively on a public blind test set, together with a transferable SSL-guided training strategy. The authors retrained the baseline models on the same synthetic dataset and used consistent STFT settings, which strengthens the fairness of the comparison. The public blind test set and the frozen SSL model also reduce the risk of circular evaluation. However, the central comparative claim rests on small AECMOS differences reported as single-run point estimates, so the statistical reliability of the headline result is not yet established.

major comments (3)
  1. [§IV-C/IV-D, Table I] The reported AECMOS scores are point estimates from a single evaluation checkpoint (Run 1668423760 Stage 0.onnx) with no confidence intervals or significance tests over the 800 blind test clips. The margins over DeepVQE-S are small: +0.07 EchoMOS in ST FE and +0.03 DegMOS in ST NE, while DeepVQE-S leads by +0.08 EchoMOS and +0.16 DegMOS in DT. Given 800 samples, the standard error of these differences could plausibly exceed the reported margins. The statements in IV-D that EchoFree 'surpasses DeepVQE-S' and in V that it achieves 'superior performance' are therefore load-bearing claims that need statistical support. I ask the authors to provide bootstrap confidence intervals over test clips, per-condition error bars, and/or results from multiple training/evaluation runs.
  2. [§IV-B, Table I] The headline complexity figure of 30 MMACs/s appears to count only the neural post-filter; the partitioned-block frequency-domain adaptive Kalman filter used in the linear stage is listed as '-' in Table I. Since the Kalman filter is part of the deployed inference pipeline, the end-to-end computational cost is not reported. To support the 'ultra lightweight' and 'superior trade-off' claims, the authors should either report the MACs of the linear filtering stage and the total system cost, or clearly state that the comparison covers only the neural component.
  3. [§IV-E, Table I] The ablation claims that the two-stage strategy yields 'superior results' compared to direct SSL loss training. However, the proposed two-stage training reduces DT EchoMOS from 3.91 (SSL-only) to 3.88, i.e., a small degradation on an important echo-cancellation condition. The paper only labels this a 'minor reduction' and does not discuss the implication. Since the two-stage method is a key contribution, the trade-off between improved ST FE/ST NE and worsened DT EchoMOS should be quantified and analyzed.
minor comments (8)
  1. [Abstract / IV-D / V] The wording is inconsistent: the Abstract says 'comparable to DeepVQE-S', while IV-D says 'surpasses DeepVQE-S' and the conclusion says 'superior performance'. Align the claims with the actual metric margins and statistical evidence.
  2. [Eq. (3)] The SSL loss formula omits the sample index and the averaging over the training batch; the text refers to 'i-th sample' but the equation only shows layer-wise summation. Clarify the notation.
  3. [Eq. (4)] The coefficient 10 multiplying the quartic term and the later 10*L_Bark in Eq. (6) mean the quartic term has an effective weight of 100. It is unclear whether this is intentional. Please clarify the role of each coefficient. Also, the phrase 'root-mean discrepancies' does not match the formula.
  4. [§III.C] The description 'first- and second-order derivatives of the first six features' is vague: which features, why only the first six, and how the 112 dimension is obtained from 100 Bark filters plus derivatives should be stated explicitly (or referenced precisely).
  5. [Table I caption] The column headers 'ST FE EchoMOS', 'ST NE DegMOS', 'DT EchoMOS', 'DT DegMOS' should be defined in the caption or text: ST FE = far-end single-talk, ST NE = near-end single-talk, DT = double-talk. This would make the table self-contained.
  6. [§IV.D] The text refers to 'Section 3.1' when describing the STFT configuration; the method is described in Section III.C. Fix the cross-reference.
  7. [General] No code is released. Given that all baselines were re-implemented/retrained by the authors, releasing the evaluation and training scripts would be important for reproducibility, especially because the headline result is based on close metric margins.
  8. [Fig. 2 caption] Typo in 'neurual' -> 'neural'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: EchoFree's central efficiency/performance claim is an external benchmark comparison, not a derivation from its own inputs.

full rationale

The paper's central claim—that EchoFree with 278K parameters and 30 MMACs is comparable to or better than low-complexity baselines—is established by evaluation on the public ICASSP 2023 AEC Challenge blind test set using the external AECMOS metric (Section IV-C, Table I). No test-set-derived quantity is used as a training target, loss weight, or fitted parameter; the model is trained on simulated data from DNS clean speech, and WavLM is frozen during training. Baselines are retrained on the same dataset, so the comparison is not a fitted-input-called-prediction situation. The self-citations to the authors' prior work ([8], [10], [18]) appear only in related-work descriptions or as motivation for the SSL loss; the benefit of the SSL loss and two-stage training is validated by the paper's own ablation study (Table I), so these citations are not load-bearing for the main claim. The Bark-scale feature design follows prior external work ([6], [14]) as a modeling choice, not as a derivation. The statistical fragility of the small AECMOS margins and the use of a single evaluation checkpoint are legitimate correctness/reproducibility concerns, but they are not circularity. The paper is self-contained against external benchmarks, so the circularity score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

No new physical or architectural entities are postulated; the contribution is a configuration of existing components. The free parameters are standard hand-set hyperparameters and loss weights, not fitted to the evaluation metric.

free parameters (4)
  • Number of Bark filters / input feature dimension = 100 filters, 112 features
    Chosen by hand following [6]; directly sets model input size and compute.
  • Loss weights in stage-2 = 10.0 (Bark), 0.5 (SSL), 0.01 (CE)
    Manual weights; stage-2 performance depends on this balance.
  • Bark gain compression coefficient c = 0.5
    Hand-set exponent in Eq. 4; the loss surface changes with c.
  • GRU hidden size and conv filter sizes = 192 units; 8/16/24/32 filters
    Architecture hyperparameters chosen without a search; not fitted to the test set.
assumptions (5)
  • domain assumption Signal model y = s + e with no separate noise term
    Equation (1); the paper explicitly ignores background noise, which limits real-world applicability.
  • domain assumption Bark-scale magnitude mask (multiplied by mixed-signal magnitude) reconstructs near-end speech with acceptable phase
    Post-processing section uses the noisy phase implicitly; standard magnitude-masking assumption.
  • domain assumption WavLM-Large frozen embeddings provide useful supervision for echo suppression
    Section III-D; relies on transfer from SSL speech representations to a front-end task.
  • domain assumption The dynamic simulation pipeline covers the ICASSP 2023 test conditions
    Section IV-A; training data is synthetic (DNS clean speech + simulated RIR/delay/nonlinearity); if the simulation distribution is unrepresentative, the benchmark result will not transfer.
  • domain assumption The adaptive Kalman filter sufficiently removes linear echo so that residual echo dominates the post-filter input
    Section III-B; post-filter is trained on estimated echo, so linear filter quality bounds what the NN can correct.

how reviews work

0 comments
Cite this review

Pith. "Pith review of EchoFree: Towards Ultra Lightweight and Efficient Neural Acoustic Echo Cancellation." pith.science (2026). https://pith.science/paper/7IVUK7SJ

@misc{pith2026250806271,
  author       = {Pith},
  title        = {Pith review of: EchoFree: Towards Ultra Lightweight and Efficient Neural Acoustic Echo Cancellation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7IVUK7SJ}},
  note         = {Machine review of arXiv:2508.06271}
}
read the original abstract

In recent years, neural networks (NNs) have been widely applied in acoustic echo cancellation (AEC). However, existing approaches struggle to meet real-world low-latency and computational requirements while maintaining performance. To address this challenge, we propose EchoFree, an ultra lightweight neural AEC framework that combines linear filtering with a neural post filter. Specifically, we design a neural post-filter operating on Bark-scale spectral features. Furthermore, we introduce a two-stage optimization strategy utilizing self-supervised learning (SSL) models to improve model performance. We evaluate our method on the blind test set of the ICASSP 2023 AEC Challenge. The results demonstrate that our model, with only 278K parameters and 30 MMACs computational complexity, outperforms existing low-complexity AEC models and achieves performance comparable to that of state-of-the-art lightweight model DeepVQE-S. The audio examples are available.

Figures

Figures reproduced from arXiv: 2508.06271 by the authors.

Figure 1
Figure 1. The hybrid approach combines a linear acoustic echo [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Neural network architecture of our proposed neurual post filter. (a) provides an overall architecture of the neural post [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The proposed two-stage training strategy: the first stage [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visualization of the output at each stage. The input audio (a) contains acoustic echo. The output of the first stage (b) [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Echo-Aware Modulation for Compact-Latent Frequency-Time Modeling in Lightweight Acoustic Echo Cancellation

    eess.AS 2026-08 conditional novelty 6.0 of 10

    An echo-aware modulation module recovers frequency-time detail in Bark-domain lightweight acoustic echo cancellation, improving quality at modest extra cost.

Reference graph

Works this paper leans on

29 extracted references · 27 canonical work pages · cited by 1 Pith paper

  1. [1]

    Proportionate normalized least-mean-squares adapta- tion in echo cancelers,

    D. L. Duttweiler, “Proportionate normalized least-mean-squares adapta- tion in echo cancelers,” IEEE Trans. Speech Audio Process. , 2000

    work page 2000

  2. [2]

    Performance analysis of unconstrained partitioned-block frequency-domain adaptive filters in under-modeling scenarios,

    Z. Luo, Z. Yu, F. Kang, F. Yang, and J. Yang, “Performance analysis of unconstrained partitioned-block frequency-domain adaptive filters in under-modeling scenarios,” EURASIP J. Adv. Signal Process. , 2024

    work page 2024

  3. [3]

    Multidelay block frequency domain adaptive filter,

    J. Soo and K. K. Pang, “Multidelay block frequency domain adaptive filter,” IEEE Trans. Acoust. Speech Signal Process. , 1990

    work page 1990

  4. [4]

    Study of the general kalman filter for echo cancellation,

    C. Paleologu, J. Benesty, and S. Ciochina, “Study of the general kalman filter for echo cancellation,” IEEE Trans. Speech Audio Process. , 2013

    work page 2013

  5. [5]

    Nonlinear residual acoustic echo suppression for high levels of harmonic distortion,

    D. A. Bendersky, J. W. Stokes, and H. S. Malvar, “Nonlinear residual acoustic echo suppression for high levels of harmonic distortion,” in Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2008, March 30 - April 4, 2008, Caesars Palace, Las V egas, Nevada, USA , 2008

    work page 2008

  6. [6]

    Acoustic echo cancellation by combining adaptive digital filter and recurrent neural network,

    L. Ma, H. Huang, P. Zhao, and T. Su, “Acoustic echo cancellation by combining adaptive digital filter and recurrent neural network,” CoRR, 2020

    work page 2020

  7. [7]

    Neuralkalman: A learnable kalman filter for acoustic echo cancellation,

    Y . Zhang, M. Yu, H. Zhang, D. Yu, and D. Wang, “Neuralkalman: A learnable kalman filter for acoustic echo cancellation,” in IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2023, Taipei, Taiwan, December 16-20, 2023 , 2023

    work page 2023

  8. [8]

    F-T-LSTM based complex network for joint acoustic echo cancellation and speech enhancement,

    S. Zhang, Y . Kong, S. Lv, Y . Hu, and L. Xie, “F-T-LSTM based complex network for joint acoustic echo cancellation and speech enhancement,” in 22nd Annual Conference of the International Speech Communication Association, Interspeech 2021, Brno, Czechia, August 30 - September 3, 2021, H. Hermansky, H. Cernock´y, L. Burget, L. Lamel, O. Scharenborg, and P....

    work page 2021

Show all 29 references
  1. [9]

    Multi-scale temporal frequency convolutional network with axial attention for multi-channel speech enhancement,

    G. Zhang, C. Wang, L. Yu, and J. Wei, “Multi-scale temporal frequency convolutional network with axial attention for multi-channel speech enhancement,” in IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2022, Virtual and Singapore, 23-27 May 2022, 2022

    2022

  2. [10]

    Multi- task deep residual echo suppression with echo-aware loss,

    S. Zhang, Z. Wang, J. Sun, Y . Fu, B. Tian, Q. Fu, and L. Xie, “Multi- task deep residual echo suppression with echo-aware loss,” in IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2022, Virtual and Singapore, 23-27 May 2022 , 2022

    2022

  3. [11]

    Multi-task sub-band network for deep residual echo suppression,

    J. Sun, D. Luo, Z. Li, J. Li, Y . Ju, and Y . Li, “Multi-task sub-band network for deep residual echo suppression,” in IEEE International Conference on Acoustics, Speech and Signal Processing ICASSP 2023, Rhodes Island, Greece, June 4-10, 2023 . IEEE, 2023

    2023

  4. [12]

    Echofilter: End- to-end neural network for acoustic echo cancellation,

    L. Ma, S. Yang, Y . Gong, X. Wang, and Z. Wu, “Echofilter: End- to-end neural network for acoustic echo cancellation,” arXiv preprint arXiv:2105.14666, 2021

    2021 arXiv

  5. [13]

    Deepvqe: Real time deep voice quality enhancement for joint acoustic echo cancellation, noise suppression and dereverberation,

    N. Ristea, E. Indenbom, A. Saabas, T. P ¨arnamaa, J. Guzvin, and R. Cutler, “Deepvqe: Real time deep voice quality enhancement for joint acoustic echo cancellation, noise suppression and dereverberation,” in 24th Annual Conference of the International Speech Communication Asso...

    2023

  6. [14]

    Low- complexity, real-time joint neural echo control and speech enhancement based on percepnet,

    J. Valin, S. V . Tenneti, K. Helwani, U. Isik, and A. Krishnaswamy, “Low- complexity, real-time joint neural echo control and speech enhancement based on percepnet,” in IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2021, Toronto, ON, Canada, ...

    2021

  7. [15]

    A hybrid approach for low-complexity joint acoustic echo and noise reduction,

    S. S. Shetu, N. K. Desiraju, J. M. M. Aponte, E. A. P. Habets, and E. Mabande, “A hybrid approach for low-complexity joint acoustic echo and noise reduction,” in 18th International Workshop on Acoustic Signal Enhancement, IWAENC 2024, Aalborg, Denmark, September 9- 12, 2024 , 2024

    2024

  8. [16]

    Align-ulcnet: Towards low-complexity and robust acoustic echo and noise reduction,

    S. S. Shetu, N. K. Desiraju, W. Mack, and E. A. P. Habets, “Align-ulcnet: Towards low-complexity and robust acoustic echo and noise reduction,” CoRR, 2024

    2024

  9. [17]

    Ultra dual- path compression for joint echo cancellation and noise suppression,

    H. Chen, J. Yu, Y . Luo, R. Gu, W. Li, Z. Lu, and C. Weng, “Ultra dual- path compression for joint echo cancellation and noise suppression,” in 24th Annual Conference of the International Speech Communication Association, Interspeech 2023, Dublin, Ireland, August 20-24, 2023 ,...

    2023

  10. [18]

    Vec-tok speech: speech vectorization and tokenization for neural speech generation,

    X. Zhu, Y . Lv, Y . Lei, T. Li, W. He, H. Zhou, H. Lu, and L. Xie, “Vec-tok speech: speech vectorization and tokenization for neural speech generation,” CoRR, vol. abs/2310.07246, 2023

    2023 arXiv

  11. [19]

    A closer look at wav2vec2 embeddings for on-device single-channel speech enhancement,

    R. Shankar, K. Tan, B. Xu, and A. Kumar, “A closer look at wav2vec2 embeddings for on-device single-channel speech enhancement,” in IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2024, Seoul, Republic of Korea, April 14-19, 2024 , 2024

    2024

  12. [20]

    State-space architecture of the partitioned-block-based acoustic echo controller,

    F. Kuech, E. Mabande, and G. Enzner, “State-space architecture of the partitioned-block-based acoustic echo controller,” in IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2014, Florence, Italy, May 4-9, 2014 , 2014

    2014

  13. [21]

    Efficient high-performance bark-scale neural network for residual echo and noise suppression,

    E. Seidel, P. Mowlaee, and T. Fingscheidt, “Efficient high-performance bark-scale neural network for residual echo and noise suppression,” in IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2024, Seoul, Republic of Korea, April 14-19, 2024 , 2024

    2024

  14. [22]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical Image Computing and Computer-Assisted Intervention - MICCAI 2015 - 18th International Conference Munich, Germany, October 5 - 9, 2015, Proceedings, Part III, ...

    2015

  15. [23]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift,

    S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in Proceedings of the 32nd International Conference on Machine Learning, ICML 2015, Lille, France, 6-11 July 2015 , F. R. Bach and D. M. Blei, Eds., 2015

    2015

  16. [24]

    Fast and accurate deep network learning by exponential linear units (elus),

    D. Clevert, T. Unterthiner, and S. Hochreiter, “Fast and accurate deep network learning by exponential linear units (elus),” in 4th International Conference on Learning Representations, ICLR 2016, San Juan, Puerto Rico, May 2-4, 2016, Conference Track Proceedings , Y . Bengio ...

    2016

  17. [25]

    Wavlm: Large-scale self-supervised pre- training for full stack speech processing,

    S. Chen, C. Wang, Z. Chen, Y . Wu, S. Liu, Z. Chen, J. Li, N. Kanda, T. Yoshioka, X. Xiao, J. Wu, L. Zhou, S. Ren, Y . Qian, Y . Qian, J. Wu, M. Zeng, X. Yu, and F. Wei, “Wavlm: Large-scale self-supervised pre- training for full stack speech processing,” 2022

    2022

  18. [26]

    ICASSP 2021 deep noise suppression challenge: Decoupling magnitude and phase optimization with a two-stage deep network,

    A. Li, W. Liu, X. Luo, C. Zheng, and X. Li, “ICASSP 2021 deep noise suppression challenge: Decoupling magnitude and phase optimization with a two-stage deep network,” in IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2021, Toronto, ON, Canada,...

    2021

  19. [27]

    Adam: A method for stochastic optimiza- tion,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimiza- tion,” in 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, Y . Bengio and Y . LeCun, Eds., 2015

    2015

  20. [28]

    ICASSP 2023 acoustic echo cancellation challenge,

    R. Cutler, A. Saabas, T. P ¨arnamaa, M. Purin, E. Indenbom, N. Ristea, J. Guzvin, H. Gamper, S. Braun, and R. Aichner, “ICASSP 2023 acoustic echo cancellation challenge,” CoRR, vol. abs/2309.12553, 2023

    2023 arXiv

  21. [29]

    AECMOS: A speech quality assessment metric for echo impairment,

    M. Purin, S. Sootla, M. Sponza, A. Saabas, and R. Cutler, “AECMOS: A speech quality assessment metric for echo impairment,” in IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2022, Virtual and Singapore, 23-27 May 2022 , 2022

    2022

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.