An adaptive filter bank based neural network approach for time delay estimation and speech enhancement

Lu Ma

REVIEW 3 major objections 4 minor 23 references

An adaptive filter bank based neural network approach for time delay estimation and speech enhancement

T0 review · 3 major / 4 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read A filter-bank neural network estimates echo delay with 91.67% accuracy and lifts mean speech quality by 0.633 PESQ.

desk verdict Novel filter-bank TDE idea, but the synthetic evaluation aligns away the real delay, so the headline numbers don't yet transfer to actual AEC. read the letter →

arxiv 2502.06098 v1 pith:GI6ZVKAQ submitted 2025-02-10 cs.SD eess.AS

Lu Ma This is my paper · ORCID

classification cs.SDeess.AS

keywords timedelayestimationacousticechocancellationadaptivefilterbankresidualsuppressionOMLSAautomaticgaincontrolneuralnetworkclassificationspeechenhancement

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

Spatial (acoustic) echo cancellation needs to know how far the far-end reference lags the microphone signal; when that delay is wrong, the adaptive filter wastes its length correcting the error instead of modeling the room. This paper proposes estimating the delay by running a bank of overlapping adaptive filters and classifying the pattern of their block-weight energies with a small neural network. On synthetic test data the estimator reaches 91.67% accuracy within a ±25 ms tolerance and 89.88% within ±5 ms, far above the open-source correlation-based AEC baseline. Around this estimator the paper builds a full echo-cancellation and speech-enhancement chain whose mean PESQ gain of 0.633 beats both the open-source AEC baseline and a much larger end-to-end neural network baseline.

What carries the argument

The load-bearing mechanism is the adaptive filter bank. Five multidelay block frequency-domain adaptive filters, each split into $N=32$ blocks with $L=8$ blocks of overlap between adjacent filters, are fed from a shared buffer of the far-end signal; the $M\times N$ block-energy values from all filters are concatenated into the input vector of a classification network taken from the voice-activity-detection path of a hybrid DSP/neural speech enhancer. The possible delays correspond one-to-one to classes, and the index of the maximal output probability is the estimated delay. This design converts a continuous cross-correlation search into a discrete pattern-recognition problem whose sensitivity to nonlinear echo and noise is governed by the learned classifier rather than by hand-tuned postprocessing.

What would settle it

Train or test the same TDE classifier with room impulse responses that are not pre-aligned to sample zero, or run it on recordings from a real room where the true echo delay is known by measurement, and compare the estimated delay against ground truth; if accuracy falls back to the correlation-baseline level, the sample-zero alignment is what carries the result.

Watch

Extended reading notes

Core claim

The paper claims that time delay estimation for acoustic echo cancellation can be recast as a classification problem: place a set of multidelay block frequency-domain adaptive filters so their time scopes overlap, compute the energy of each filter's weight blocks, concatenate those energies, and let a lightweight feed-forward classification network output the most probable delay. Because the filter bank itself performs linear echo estimation, its weight pattern already encodes where the echo sits; the classifier only has to read that pattern. The same best filter output then feeds a neural residual-echo suppressor followed by OMLSA-based denoising that treats residual echo as a special noise, and a sigmoid-smoothed automatic gain control. The paper reports that this combined system improves mean perceptual speech quality by 0.633 PESQ over the unprocessed signal, with a model size of 794K parameters, better than a 1827K-parameter end-to-end neural baseline and the open-source AEC baseline.

Load-bearing premise

The training data assumes every room impulse response has its direct path shifted to sample zero, so the only delay the system must learn is the one artificially added during synthesis; real acoustic echo paths have unknown and varying direct-path delays that this training does not cover.

Editorial extensions

If this is right

A correlation-free TDE that loses only 1.79 percentage points when the tolerance tightens from $\pm 25$ ms to $\pm 5$ ms can support high-precision adaptive filtering without extra filter length.
The full AEC chain, at 794K parameters, outperforms an end-to-end neural AEC model more than twice its size in mean PESQ gain, so the linear-filter-first architecture is parameter-efficient.
Treating residual echo as a special noise within OMLSA reduces spectral distortion, lowering the variance of PESQ gain compared to neural suppression alone.
The sigmoid gain-smoothing rule lets the automatic gain control raise gain quickly without introducing frame-boundary discontinuities in the spectrum.

Reading between the lines

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

Editorial inference: because every RIR was pre-aligned with its direct path at sample zero, the reported TDE accuracy is probably an upper bound for real rooms, where the direct-path delay is unknown and must itself be discovered.
Editorial inference: the block-energy pattern of an adaptive filter bank may be a general, low-cost feature for any task that needs to localize a delayed copy of a reference signal, not just echo cancellation.
Editorial inference: one could test transfer directly by training the classifier on unaligned RIRs and measuring how much accuracy drops; that drop quantifies how much of the reported performance comes from the alignment shortcut.

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.

Desk Editor's Note

read the letter

Quick take: the adaptive-filter-bank delay estimator is a genuinely new twist, and the system is assembled from sensible components, but the evaluation currently aligns the RIR direct path to zero, which removes the exact delay a real TDE would have to find—so the 91.67% accuracy and 0.633 PESQ gain are measured on an artificially clean version of the task.

What's actually new: using the block-energy profile of a bank of overlapping adaptive filters as input features to a small classifier for delay estimation. That is not in the prior TDE literature I know. The idea is cheap, plausible, and the paper honestly credits WebRTC AEC3 and RNNoise for the architecture. The AGC smoothing with a sigmoid cross-fade is also a reasonable engineering touch. The paper is clearly written and follows a sensible train/eval split.

Soft spots: first, the RIR alignment issue. Section 3.1 says the direct path of every RIR is identified and set to position 0. That means the only delay the classifier sees is the artificial τ added to the mixture. In a real AEC, the reference-to-microphone delay includes the loudspeaker-to-microphone propagation plus hardware latency, and it varies. No evidence here that the filter-bank features capture that unknown delay. Table 1 therefore describes a synthetic subproblem, not the problem AEC needs solved. Second, the internal numbers don't add up: N=32, M=5 gives M×N=160, not 532, and the category count M×N−L is also inconsistent. Third, no significance tests, no error bars, and the PESQ variance in Table 2 is large relative to the mean gains. Fourth, the suppression network is trained on the same generated distribution and uses the estimated delay to create the error signal, so the AEC gains are also in-distribution.

None of this means the method is wrong. The measured numbers are probably correct on the defined synthetic task. But the paper overclaims robustness, and the missing pieces—public code, a real RIR delay setup, or at least an evaluation with unaligned RIRs—are essential before the results mean much for practice.

My recommendation: send it to peer review, but clearly flag that the TDE evaluation needs to be redone with unaligned RIRs or real recordings, and the arithmetic errors need fixing. A competent referee can sort this out in one round. I wouldn't cite it in my own work until that verification exists.

Referee Report

3 major / 4 minor

Summary. The paper proposes a hybrid AEC system with four components: (i) a time-delay estimator built from a bank of five multidelay block adaptive filters whose per-block weight energies are classified by a lightweight neural network; (ii) a residual-echo and noise suppression network based on RNNoise-style MFCC features with a double-talk detection output; (iii) an OMLSA postfilter that treats residual echo as a special noise; and (iv) an automatic gain control with sigmoid-based gain smoothing. Experiments on a synthetic dataset derived from the AEC Challenge 2021 and DNS-Challenge corpora report TDE accuracies of 91.67 percent within 25 ms and 89.88 percent within 5 ms, and a mean PESQ gain of 0.633 for the complete processing chain, outperforming the reported WebRTC and DTLN baselines.

Significance. The central idea of using the energy pattern of an adaptive-filter bank as a feature for delay classification is plausible and well aligned with current hybrid DSP/neural AEC research. The paper also provides useful implementation details, including filter-bank geometry, training data generation, model sizes, and comparisons with representative baselines. If the reported TDE accuracy and PESQ gains were shown to transfer to realistic acoustic echo paths, the contribution would be solid. However, the empirical evidence as presented is entirely in-distribution on a synthetic pipeline in which the direct-path delay of every room impulse response is removed; this is a load-bearing gap for the paper's main claim. The lack of statistical testing and the arithmetic inconsistency in the network dimensions further reduce confidence.

major comments (3)

[3.1] Section 3.1 states that for every room impulse response 'the begin of the direct path was identified and set to position 0'. The TDE training data are then created by delaying the resulting microphone mixture by an artificially chosen tau. This means the only delay present in the adaptive-filter features is the imposed uniform shift; the physical direct-path delay of the echo path, which is precisely the quantity a real time-delay estimator must handle, is absent. Table 1 and Table 2 are computed on this aligned synthetic pipeline, so the reported accuracy and PESQ gains do not support the conclusion that the method is robust for real TDE/AEC. The authors should add experiments on unaligned RIRs with varied direct-path delays, or on real recorded data, and report the corresponding accuracies and PESQ gains.
[2.1] In Section 2.1, the stated parameters N=32, M=5 give M x N = 160, not 532 as printed. With L=8, the number of buffer positions and classifier categories is M x N - L = 152, not 155. Since the text relies on these numbers to define the classifier input dimension and output categories, the arithmetic error makes the architecture ambiguous. Please correct the values and state the exact input and output dimensions used in the implementation.
[3.2] The evaluation reports point estimates without confidence intervals or significance tests. In Table 2, the PESQ-gain variance values (0.101 for +++OMLSA, 0.162 for ++NN) are large relative to the differences in mean gains between the proposed scheme (0.633) and the End2End baseline (0.545). Without the number of test clips, error bars, or paired statistical tests, the claimed superiority over the baselines is not established. Please provide these statistics or soften the comparative claims accordingly.

minor comments (4)

[3.1] The text says the TDE data are prepared 'as is illustrated in Fig. 3(a)', but the correct reference is Fig. 2(a); the same inconsistency affects the figure reference for the suppression network data preparation in Section 2.2.
[2.2] The notation for the speech-present probability is inconsistent between Eqs. (2) and (3): it appears as p'(k,l) and p(k,l), and also as p(l,k). Please unify the notation and define each symbol at first use.
[2.1] The sentence 'As is is shown in Fig. 1' contains a duplicated word, and Section 3.2 refers to 'OLMSA' instead of 'OMLSA'.
[5] The text states that all components can be realized from open-sourced code, but no code, trained models, or configuration files are released. A statement about the planned release would strengthen reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No derivation-equivalent circularity: the delay classifier and suppression network are trained and tested on held-out synthetic data, and no prediction reduces by construction to its fitted input.

full rationale

The paper's TDE is a supervised classifier: adaptive-filter weight energies from a bank of multidelay filters are used as features, and the ground-truth label is the artificially inserted delay τ(n). The reported accuracies in Table 1 are empirical held-out results from the same generative pipeline, not identities forced by the definition of the features. Likewise, the AEC evaluation in Table 2 compares learned masks, OMLSA gains, and AGC smoothing on held-out synthetic clips. The main validity limitation is external, not circular: Section 3.1 aligns every RIR direct path to sample 0 and then injects delay as a global uniform shift of the noisy mixture, so the model only learns to recover that injected shift and real acoustic-path delays are not tested. That is a missing-evidence concern, not a self-referential derivation. The only self-citation, reference [7], is paired with independent works [8,9] to motivate the joint linear-filtering-plus-residual-suppression paradigm, and it is not used to forbid alternatives or to import a uniqueness theorem. No equation in the paper sets the predicted delay equal to the training label by construction, and no fitted parameter is renamed as a prediction. Therefore the derivation chain is not circular.

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

The paper relies on several manually chosen constants (filter bank geometry, OMLSA and AGC parameters, data augmentation ranges) and on domain assumptions about RIR alignment and nonlinear echo modeling. No independent evidence is given that these choices generalize beyond the synthetic setup.

free parameters (5)

Filter bank geometry: N=32 blocks per filter, L=8 block overlap, M=5 filters = N=32, L=8, M=5
Section 2.1 sets these values without justification; they determine delay resolution and maximum estimable delay.
OMLSA noise decision threshold and smoothing factor alpha_p = threshold=0.5, alpha_p unspecified
Equation (1)-(2) use a fixed 0.5 threshold and an unstated smoothing factor alpha_p to classify noise frequencies.
AGC target levels Amean_def and Amax_def, gain smoothing alpha, saturation guard bGk, sigmoid range -5.0 to 10.0 = not reported
Section 2.3 defines these constants for gain control, but no values or tuning procedure are given.
Training data augmentation ranges: SER -30 to 30 dB, SNR -10 to 30 dB, both in 5 dB steps; delay 0-500 ms in 10 ms steps = as stated
Section 3.1 generates 2e5 TDE clips and 4e5 AEC clips using these ranges; the choice directly shapes what the networks learn.
DTD/VAD thresholds and MFCC feature configuration = unspecified
Section 2.2 mentions thresholds for voice activity and double talk detection but does not report their values or how they were chosen.

assumptions (4)

domain assumption The direct path of each room impulse response is aligned to sample position 0, so the only time delay is the artificially added one.
Section 3.1 states 'the begin of the direct path was identified and set to position 0'. This removes real RIR onset uncertainty from the learning problem.
domain assumption The nonlinear echo model from [12] adequately represents real nonlinear loudspeaker behavior.
Section 3.1 uses this model to synthesize nonlinear echo for training and evaluation, without validating it against measured nonlinear devices.
ad hoc to paper Block filter energies are a sufficient feature set for delay classification.
Section 2.1 assumes that the concatenated energy profile uniquely encodes the delay under the training conditions; no convergence analysis of the adaptive filters is provided.
domain assumption The OMLSA algorithm from [15] remains valid when residual echo is treated as a special noise.
Section 2.2 directly adopts OMLSA without re-deriving its statistical model for the echo-as-noise case.

how reviews work

0 comments

Cite this review

Pith. "Pith review of An adaptive filter bank based neural network approach for time delay estimation and speech enhancement." pith.science (2026). https://pith.science/paper/GI6ZVKAQ

@misc{pith2026250206098,
  author       = {Pith},
  title        = {Pith review of: An adaptive filter bank based neural network approach for time delay estimation and speech enhancement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GI6ZVKAQ}},
  note         = {Machine review of arXiv:2502.06098}
}

read the original abstract

Time delay estimation (TDE) plays a key role in acoustic echo cancellation (AEC) using adaptive filter method. Considerable residual echo will be left if estimation error arises. Here, in this paper, we proposed an adaptive filter bank based neural network approach where the delay is estimated by a bank of adaptive filters with overlapped time scope, and all the energy of filter weights are concatenated and feed to a classification network. The index with maximal probability is chosen as the estimated delay. Based on this TDE, an AEC scheme is designed using a neural network for residual echo and noise suppression, and the optimally-modified log-spectral amplitude (OMLSA) algorithm is adopted to make it robust. Also, a robust automatic gain control (AGC) scheme with spectrum smoothing method is designed to amplify speech segments. Performance evaluations reveal that higher performance can be achieved for our scheme.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 23 canonical work pages

[1]

INTRODUCTION Acoustic echo arises when the microphone at the near-end picks up the loudspeaker’s sound plus its reverberation and heard by the speaker itself at the far-end. Since there is a reference signal named the far-end, adaptive filter (AF) is al- ways employed for acoustic echo cancellation (AEC) [1, 2]., i.e., the echo is estimated by adaptive fi...

work page Pith review arXiv 2025
[2]

FRAMEWORK The framework is illustrated in Fig. 1. Three modules are in- cluded, i.e., time delay estimation, residual echo suppression and automatic gain control. The time delay is first estimated by the adaptive filter banks with classification neural network, then the optimal filter output is selected and used for resid- ual echo suppression with neural...

work page
[3]

Dataset Preparation The dataset of AEC challenge 2021 is utilized for AEC model training

EXPERIMENTS 3.1. Dataset Preparation The dataset of AEC challenge 2021 is utilized for AEC model training. The preparation is referred to in [14]. Only the far-end and the echo signals are used and cut into chunks of 4 s. Clean speech and noise corpus from DNS-Challenge 2021 [17] are used. The near-end signal is convolved with RIR randomly selected from [...

work page 2021
[4]

CONCLUSION Inspired by WebRTC AEC3, a time delay estimation scheme based on a bank of adaptive filters and a lightweight classifi- cation neural network was proposed, resulting in higher accu- racy and more robust in term of time delay estimation. This TDE was then employed for AEC together with a suppression neural network for residual echo and noise sup...

work page
[5]

On the implementation of a partitioned block frequency do- main adaptive filter (pbfdaf) for long acoustic echo can- cellation,

Jos ´eM. P ´aez Borrallo and Mariano Garcia Otero, “On the implementation of a partitioned block frequency do- main adaptive filter (pbfdaf) for long acoustic echo can- cellation,” Signal Processing, vol. 27, no. 3, pp. 301– 315, 1992, Acoustic Echo Control

work page 1992
[6]

Multidelay block frequency domain adaptive filter,

J.-S. Soo and K.K. Pang, “Multidelay block frequency domain adaptive filter,” IEEE Transactions on Acous- tics, Speech, and Signal Processing , vol. 38, no. 2, pp. 373–376, 1990

work page 1990
[7]

Adapta- tion of a memoryless preprocessor for nonlinear acous- tic echo cancelling,

Alexander Stenger and Walter Kellermann, “Adapta- tion of a memoryless preprocessor for nonlinear acous- tic echo cancelling,” Signal Processing, vol. 80, no. 9, pp. 1747–1760, 2000

work page 2000
[8]

Nonlin- ear acoustic echo cancellation using adaptive orthogo- nalized power filters,

F. Kuech, A. Mitnacht, and W. Kellermann, “Nonlin- ear acoustic echo cancellation using adaptive orthogo- nalized power filters,” in Proceedings. (ICASSP ’05). IEEE International Conference on Acoustics, Speech, and Signal Processing, 2005. , 2005, vol. 3, pp. iii/105– iii/108 V ol. 3

work page 2005

Show all 23 references

[9]

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

Diego A. Bendersky, Jack W. Stokes, and Henrique S. Malvar, “Nonlinear residual acoustic echo suppression for high levels of harmonic distortion,” in2008 IEEE In- ternational Conference on Acoustics, Speech and Signal Processing, 2008, pp. 261–264

2008
[10]

Spectral feature-based nonlinear residual echo suppression,

Andreas Schwarz, Christian Hofmann, and Walter Kellermann, “Spectral feature-based nonlinear residual echo suppression,” in 2013 IEEE Workshop on Applica- tions of Signal Processing to Audio and Acoustics, 2013, pp. 1–4

2013
[11]

Acoustic echo cancellation by combining adaptive dig- ital filter and recurrent neural network,

Lu Ma, Hua Huang, Pei Zhao, and Tengrong Su, “Acoustic echo cancellation by combining adaptive dig- ital filter and recurrent neural network,” arXiv preprint arXiv:2005.09237, 2020

2005 arXiv
[12]

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

Jean-Marc Valin, Srikanth Tenneti, Karim Helwani, Umut Isik, and Arvindh Krishnaswamy, “Low- complexity, real-time joint neural echo control and speech enhancement based on percepnet,” in ICASSP 2021-2021 IEEE International Conference on Acous- tics, Speech and Signal Processi...

2021
[13]

Nn3a: Neural network supported acoustic echo can- cellation, noise suppression and automatic gain control for real-time communications,

Ziteng Wang, Yueyue Na, Biao Tian, and Qiang Fu, “Nn3a: Neural network supported acoustic echo can- cellation, noise suppression and automatic gain control for real-time communications,” in ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Proces...

2022
[14]

Cad-aec: Context-aware deep acoustic echo cancel- lation,

Amin Fazel, Mostafa El-Khamy, and Jungwon Lee, “Cad-aec: Context-aware deep acoustic echo cancel- lation,” in ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020, pp. 6919–6923

2020
[15]

A neural network-based nonlinear acous- tic echo canceller,

Mhd Modar Halimeh, Christian Huemmer, and Walter Kellermann, “A neural network-based nonlinear acous- tic echo canceller,” IEEE Signal Processing Letters, vol. 26, no. 12, pp. 1827–1831, 2019

2019
[16]

Deep learning for acous- tic echo cancellation in noisy and double-talk scenar- ios,

H. Zhang and Deliang Wang, “Deep learning for acous- tic echo cancellation in noisy and double-talk scenar- ios,” in INTERSPEECH, 2018

2018
[17]

Deep multitask acoustic echo cancellation.,

Amin Fazel, Mostafa El-Khamy, and Jungwon Lee, “Deep multitask acoustic echo cancellation.,” in Inter- speech, 2019, pp. 4250–4254

2019
[18]

Acoustic echo cancellation with the dual-signal transformation lstm network,

Nils L Westhausen and Bernd T Meyer, “Acoustic echo cancellation with the dual-signal transformation lstm network,” in ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2021, pp. 7138–7142

2021
[19]

Speech enhance- ment for non-stationary noise environments,

Israel Cohen and Baruch Berdugo, “Speech enhance- ment for non-stationary noise environments,” Signal processing, vol. 81, no. 11, pp. 2403–2418, 2001

2001
[20]

A hybrid dsp/deep learning approach to real-time full-band speech enhancement,

Jean-Marc Valin, “A hybrid dsp/deep learning approach to real-time full-band speech enhancement,” in 2018 IEEE 20th international workshop on multimedia signal processing (MMSP). IEEE, 2018, pp. 1–5

2018
[21]

Icassp 2021 deep noise suppression challenge,

Chandan KA Reddy, Harishchandra Dubey, Vishak Gopal, Ross Cutler, Sebastian Braun, Hannes Gam- per, Robert Aichner, and Sriram Srinivasan, “Icassp 2021 deep noise suppression challenge,” in ICASSP 2021-2021 IEEE International Conference on Acous- tics, Speech and Signal Proces...

2021
[22]

A study on data augmentation of reverberant speech for robust speech recognition,

Tom Ko, Vijayaditya Peddinti, Daniel Povey, Michael L Seltzer, and Sanjeev Khudanpur, “A study on data augmentation of reverberant speech for robust speech recognition,” in 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2017, pp....

2017
[23]

Poconet: Better speech enhancement with frequency-positional embeddings, semi-supervised con- versational data, and biased loss,

Umut Isik, Ritwik Giri, Neerad Phansalkar, Jean- Marc Valin, Karim Helwani, and Arvindh Krish- naswamy, “Poconet: Better speech enhancement with frequency-positional embeddings, semi-supervised con- versational data, and biased loss,” arXiv preprint arXiv:2008.04470, 2020

2008 arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.

[1] [1]

INTRODUCTION Acoustic echo arises when the microphone at the near-end picks up the loudspeaker’s sound plus its reverberation and heard by the speaker itself at the far-end. Since there is a reference signal named the far-end, adaptive filter (AF) is al- ways employed for acoustic echo cancellation (AEC) [1, 2]., i.e., the echo is estimated by adaptive fi...

work page Pith review arXiv 2025

[2] [2]

FRAMEWORK The framework is illustrated in Fig. 1. Three modules are in- cluded, i.e., time delay estimation, residual echo suppression and automatic gain control. The time delay is first estimated by the adaptive filter banks with classification neural network, then the optimal filter output is selected and used for resid- ual echo suppression with neural...

work page

[3] [3]

Dataset Preparation The dataset of AEC challenge 2021 is utilized for AEC model training

EXPERIMENTS 3.1. Dataset Preparation The dataset of AEC challenge 2021 is utilized for AEC model training. The preparation is referred to in [14]. Only the far-end and the echo signals are used and cut into chunks of 4 s. Clean speech and noise corpus from DNS-Challenge 2021 [17] are used. The near-end signal is convolved with RIR randomly selected from [...

work page 2021

[4] [4]

CONCLUSION Inspired by WebRTC AEC3, a time delay estimation scheme based on a bank of adaptive filters and a lightweight classifi- cation neural network was proposed, resulting in higher accu- racy and more robust in term of time delay estimation. This TDE was then employed for AEC together with a suppression neural network for residual echo and noise sup...

work page

[5] [5]

On the implementation of a partitioned block frequency do- main adaptive filter (pbfdaf) for long acoustic echo can- cellation,

Jos ´eM. P ´aez Borrallo and Mariano Garcia Otero, “On the implementation of a partitioned block frequency do- main adaptive filter (pbfdaf) for long acoustic echo can- cellation,” Signal Processing, vol. 27, no. 3, pp. 301– 315, 1992, Acoustic Echo Control

work page 1992

[6] [6]

Multidelay block frequency domain adaptive filter,

J.-S. Soo and K.K. Pang, “Multidelay block frequency domain adaptive filter,” IEEE Transactions on Acous- tics, Speech, and Signal Processing , vol. 38, no. 2, pp. 373–376, 1990

work page 1990

[7] [7]

Adapta- tion of a memoryless preprocessor for nonlinear acous- tic echo cancelling,

Alexander Stenger and Walter Kellermann, “Adapta- tion of a memoryless preprocessor for nonlinear acous- tic echo cancelling,” Signal Processing, vol. 80, no. 9, pp. 1747–1760, 2000

work page 2000

[8] [8]

Nonlin- ear acoustic echo cancellation using adaptive orthogo- nalized power filters,

F. Kuech, A. Mitnacht, and W. Kellermann, “Nonlin- ear acoustic echo cancellation using adaptive orthogo- nalized power filters,” in Proceedings. (ICASSP ’05). IEEE International Conference on Acoustics, Speech, and Signal Processing, 2005. , 2005, vol. 3, pp. iii/105– iii/108 V ol. 3

work page 2005