Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Learning Robust Spatial Representations from Binaural Audio through Feature Distillation

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper claims that predicting clean binaural spatial features from noisy audio during pretraining produces encoder weights that, after fine-tuning, beat fully supervised models for direction-of-arrival estimation in noise.

desk verdict A promising binaural SSL pretraining idea with credible simulated results, but it needs a generic pretraining control and comparisons to SSLSAR/IPDNet to prove the spatial targets are doing the work. read the letter →

arxiv 2508.20914 v1 pith:QKTF74GK submitted 2025-08-28 cs.SD cs.LGeess.AS

Holger Severin Bovbjerg (1) , Jan {O}stergaard (1) , Jesper Jensen (1 , 2) , Shinji Watanabe (3) , Zheng-Hua Tan ((1) Aalborg University (2) Eriksholm Research Centre , (3) Carnegie Mellon University) This is my paper · ORCID
classification cs.SDcs.LGeess.AS
keywords binauralaudiodirection-of-arrivalestimationfeaturedistillationself-supervisedlearningspatialrepresentationinterauralcuesnoiserobustnessConformerencoder
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

This paper proposes a pretraining recipe for binaural audio that learns where a sound is coming from without needing any direction labels. The idea is to compute classic spatial cues—timing and level differences between the ears—from clean speech, then train a neural encoder to reproduce those cues from noisy, reverberant versions of the same audio. After pretraining, the cue-prediction head is discarded and the encoder is fine-tuned for direction-of-arrival estimation. On a simulated binaural benchmark, the pretrained models beat both fully supervised neural baselines and classic cross-correlation methods, with the best variant reducing average angular error by 46.44% relative to the strongest supervised baseline. The advantage persists with as little as ten minutes of labelled data, which matters because binaural direction labels are scarce.

What carries the argument

Key machinery: Spatial Feature Distillation (SFD). A target extractor computes a per-frame spatial feature vector—for example, phase-transformed cross-correlation or interaural phase difference plus level difference—from a clean binaural recording; a causal Conformer encoder, a transformer with convolutional modules constrained to see only past frames, processes the noisy and reverberant STFT input; a linear head predicts the target; training minimizes mean squared error. After pretraining, the linear head is thrown away and the encoder weights initialize a direction-of-arrival classification model. The causal mask makes the representation usable for frame-level streaming processing, which i

What would settle it

Run the same pretraining with targets computed from noisy and reverberant speech instead of clean speech. If fine-tuned direction-of-arrival accuracy stays at the same level, the gains are not caused by clean-target distillation; if the gains disappear, the clean-target assumption is essential. Separately, evaluate the fine-tuned models on real binaural recordings from a head-and-torso simulator outside the simulated HRTF and RIR setup; a collapse in accuracy would show the result depends on simulation.

Watch

Extended reading notes

Core claim

The central claim is that classic spatial features computed from clean binaural speech make effective training targets for a representation of sound-source location, even though no direction labels are used during pretraining. The paper's Spatial Feature Distillation framework runs a causal encoder on STFT features of noisy and reverberant binaural input and trains it to match target features computed from the corresponding clean signal: generalized cross-correlation, its phase-transformed version, cross-power-spectrum phase, or interaural level plus phase differences. The learned encoder weights are then copied into a direction-of-arrival classifier and fine-tuned with cross-entropy on quan

Load-bearing premise

The framework depends on having clean, non-reverberant binaural speech to compute the target features; if no clean reference is available, the pretraining stage cannot be run as designed, and the simulation-to-real transfer is unproven.

Editorial extensions

If this is right

  • Direction-of-arrival estimation can be improved without collecting new direction labels: any large corpus of unlabelled binaural speech, simulated or real, can drive the pretraining step.
  • With only ten minutes of labelled data, fine-tuned pretrained models stay accurate in noise, while supervised baselines degrade sharply; this lowers the labelling cost for new acoustic environments.
  • The choice of spatial target matters: phase-based targets transfer much better than a target that includes interaural level difference, so future systems should focus on phase cues.
  • Because the encoder is causal and frame-level, the learned representation can be plugged into streaming direction-of-arrival systems for hearing aids and teleconferencing, not just offline analysis.

Reading between the lines

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

  • An implicit extension is to replace the clean-speech target with a model-based target computed from a simulated acoustic model, which would remove the clean-speech assumption in settings where only noisy recordings exist.
  • The same distillation recipe could be applied to other spatial attributes, such as interaural coherence or binaural room impulse responses, yielding pretrained representations for tasks beyond direction-of-arrival, including speech separation and spatial audio coding.
  • The finding that adding interaural level difference hurts performance suggests the model may be latching onto level cues that are less invariant to noise; testing with level-normalized or temporally smoothed level-difference targets would tell whether the cue itself or its representation is the problem.
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 Spatial Feature Distillation (SFD), a pretraining method for binaural audio. During pretraining, a causal Conformer encoder is trained to predict spatial features (GCC, GCC-PHAT, CPS phase, or ILD+IPD) computed from clean binaural speech, given noisy and reverberant input. The feature predictor is then discarded and the encoder is fine-tuned for direction-of-arrival (DoA) classification. Experiments on a simulated binaural LibriSpeech/LibriLight benchmark with ARI HRTFs and SLR28 RIRs report that SFD-pretrained models achieve lower mean angular error than supervised DNN baselines and GCC-PHAT-argmax, especially at low SNRs and with small amounts of labelled data.

Significance. If the central claim holds, SFD is a useful contribution to a relatively underexplored area: self-supervised/unsupervised spatial representation learning for binaural audio. The paper releases code and data, evaluates four spatial feature targets, and shows a large practical gain in low-labelled-data regimes. However, the claim that the spatial nature of the pretraining targets is responsible for the gains is not isolated by the current experimental design, and the evaluation omits the most relevant spatial SSL baselines. These issues do not invalidate the empirical results but they do affect what can be concluded from them.

major comments (4)
  1. [§4.2, §5.1, Table 2] There is no control for generic denoising pretraining. SFD models are pretrained on 960 h of unlabelled speech with a clean-target prediction loss (Eq. 14), while the supervised baselines are trained from scratch on 1 h of labels. The reported gains, e.g. SFD-CPSPhase 7.05° vs. GCC-PHAT-DNN 13.16°, could therefore be due to the large unlabelled pretraining corpus and a denoising objective rather than to the spatial nature of the targets. Please add a control that pretrains the same encoder with the same data augmentation and fine-tuning pipeline but predicts a non-spatial clean target, such as clean STFT real/imaginary parts or clean magnitude, and compare downstream DoA MAE. Without this control, the central contribution — spatial feature distillation — is not separately validated.
  2. [§4.4, §5.1] The evaluation does not compare against the most relevant prior work, SSLSAR [28] and IPDNet [16], even though both are cited and discussed. SSLSAR is a binaural self-supervised method and IPDNet is a direct-path IPD enhancement method; these are closer to SFD in spirit than the fully-supervised baselines. Without such comparisons, the claim that SFD improves over spatial-representation-learning baselines is not established. At minimum, a discussion of why these baselines cannot be directly compared, or an experimental comparison, is needed.
  3. [§5.1, Table 2] The paper reports only a single training run per configuration. The bootstrap standard errors in Table 2 quantify test-frame resampling, not model training variance. The text says pretrained models show 'significant improvements', but no statistical test across seeds is provided. Given the small fine-tuning datasets and stochastic training, please report results over at least 3–5 seeds, or temper the significance language accordingly.
  4. [§4.1, §6] The pretraining framework assumes access to clean binaural speech to compute spatial targets. The paper acknowledges this in Section 6, but it is a load-bearing limitation: in many real applications, clean paired binaural signals are not available. Moreover, the phrase 'without the need for data labels' in the abstract is accurate only in the sense of spatial annotations; clean speech is itself a strong oracle signal. Please state this limitation in the abstract or introduction and discuss which real-world settings can provide the required clean targets.
minor comments (5)
  1. [§1] Typos: 'algiorithms' should be 'algorithms'; 'revereberant' should be 'reverberant'; 'a the concatenation' should be 'the concatenation'.
  2. [§5.1] The text says 'models trained on the Binaural LibriLight 10h dataset', but Table 2's caption states models are trained on 1h. Clarify which is correct.
  3. [§3, Eq. (12)] The notation X' in Eq. (12) is not defined; it appears to denote clean STFT features, while the noisy branch uses a different notation. Please define primes explicitly.
  4. [§4.2, Table 1] The target 'CPS-PHAT phase' is not clearly defined in the text. Equation (5) defines IPD as the phase of the CPS; explain how 'CPS-PHAT phase' differs from IPD, or use consistent terminology.
  5. [§5, Figure 2] Figure 2 would benefit from explicit axis labels for the three subpanels and a note on whether the plotted MAE is averaged over all noise types, as in Table 2.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: pretraining targets are fixed functions of clean binaural signals and downstream DoA evaluation uses separate labels and fine-tuning.

full rationale

The paper's derivation chain is an empirical pipeline, not a closed-form derivation. The pretraining targets z[n]=SFE(X'_{m1}, X'_{m2}) are deterministic functions of clean binaural signals (GCC, GCC-PHAT, CPS phase, ILD+IPD); they are not fitted parameters and do not encode the downstream DoA labels. The predictor is discarded after pretraining, and the encoder is fine-tuned on separate DoA cross-entropy labels, so no fitted input is renamed as a prediction. Citations to prior work by the same author group (e.g., [40] for noise types, [37] for simulated binaural data) are data-selection references, not load-bearing justifications of the central claim. The paper explicitly acknowledges the clean-speech assumption as a limitation (Section 6), and the simulated-HRTF generalization issue (Section 4.1); these affect applicability and robustness, not circularity. The absence of a generic denoising-pretraining control could weaken attribution of the gains to spatial targets, but that is a confound/completeness concern, not a reduction of the result to its inputs by construction. The empirical benchmark against supervised and classic baselines is self-contained, so circularity is low.

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

The method introduces no new physical entities or formal postulates. It relies on existing spatial features and standard binaural simulation practices. The main hidden costs are the hand-chosen hyperparameters and the domain assumptions of simulation fidelity and clean-speech availability, which the paper partially acknowledges.

free parameters (4)
  • STFT parameters (window 400, hop 160, FFT 512)
    Chosen by the authors for the encoder input; the paper reports no sensitivity analysis, so the results are conditional on these settings.
  • Conformer hyperparameters (2 layers, 64-dim embedding, 4 heads, kernel 31)
    Taken from [28] with a smaller embedding; no ablation of these choices is reported.
  • DoA classification resolution (5 degrees) = 5 degrees
    Defines the classification task and thus the achievable MAE ceiling; finer resolution would change the numbers.
  • Noise augmentation SNR range = [-20, 20] dB
    Uniform sampling over this range during pretraining and fine-tuning; the robustness results are specific to this range.
assumptions (4)
  • domain assumption Binaural simulation using ARI HRTFs and SLR28 RIRs faithfully represents real binaural acoustic conditions.
    Section 4.1 states the dataset is simulated and notes this 'may limit real-world generalization'.
  • domain assumption Classic spatial features (GCC, IPD, ILD) computed from clean speech are valid and sufficient targets for learning DoA-relevant representations.
    This is the core premise of the SFD framework, described in Section 3.
  • domain assumption A causal 2-layer Conformer with STFT input can learn the mapping from noisy/reverberant audio to clean spatial features.
    The architecture in Section 4.2 is assumed capable of the pretraining task; no capacity analysis is given.
  • domain assumption Single static source, horizontal plane, fixed elevation and azimuth in [-90, 90] covers the target application space.
    Section 4.1 fixes the scenario; the conclusion lists extension to moving sources and multiple speakers as future work.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Robust Spatial Representations from Binaural Audio through Feature Distillation." pith.science (2026). https://pith.science/paper/QKTF74GK

@misc{pith2026250820914,
  author       = {Pith},
  title        = {Pith review of: Learning Robust Spatial Representations from Binaural Audio through Feature Distillation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QKTF74GK}},
  note         = {Machine review of arXiv:2508.20914}
}
read the original abstract

Recently, deep representation learning has shown strong performance in multiple audio tasks. However, its use for learning spatial representations from multichannel audio is underexplored. We investigate the use of a pretraining stage based on feature distillation to learn a robust spatial representation of binaural speech without the need for data labels. In this framework, spatial features are computed from clean binaural speech samples to form prediction labels. These clean features are then predicted from corresponding augmented speech using a neural network. After pretraining, we throw away the spatial feature predictor and use the learned encoder weights to initialize a DoA estimation model which we fine-tune for DoA estimation. Our experiments demonstrate that the pretrained models show improved performance in noisy and reverberant environments after fine-tuning for direction-of-arrival estimation, when compared to fully supervised models and classic signal processing methods.

Figures

Figures reproduced from arXiv: 2508.20914 by the authors.

Figure 1
Figure 1. Spatial Feature Distillation framework overview. effective training targets for learning robust spatial features, removing the need for DoA labels. Specifically, we propose a pretraining task where a DNN predicts spatial features, computed from a clean microphone signal, from a noisy and reverberant input, similar to some noise-robust SSL methods [21], [24], [25]. We refer to this framework as Spatial Feature Distil… view at source ↗
Figure 2
Figure 2. Comparison of mean angular error (MAE) at different SNR levels, varying the amount of training data. 10h dataset. After fine-tuning for DoA prediction, all pretrained models show significant improvements in test set performance over reference methods. The SFD-CPSPhase model achieves the lowest MAE at all noise levels, outperforming the best supervised model (GCC￾PHAT-DNN) by 46.44 % on average. The SFD-GCC-PHAT mode… view at source ↗

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. Systematic Evaluation of Time-Frequency Features for Binaural Sound Source Localization

    eess.AS 2025-11 conditional novelty 6.0 of 10

    For binaural sound localization, ILD+IPD suffices on matched speech, while channel phase spectrograms plus ILD and IPD generalize best to out-of-domain sounds, and feature choice matters more than model size.

Reference graph

Works this paper leans on

41 extracted references · 37 canonical work pages · cited by 1 Pith paper

  1. [28]

    Self-supervised learning of spatial acoustic representation with cross-channel signal reconstruction and multi-channel conformer,

    B. Yang and X. Li, “Self-supervised learning of spatial acoustic representation with cross-channel signal reconstruction and multi-channel conformer,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 32, pp. 4211–4225, 2024

    work page 2024

  2. [16]

    Ipdnet: A universal direct-path ipd esti- mation network for sound source localization,

    Y . Wang, B. Yang, and X. Li, “Ipdnet: A universal direct-path ipd esti- mation network for sound source localization,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 32, pp. 5051–5064, 2024

    work page 2024

  3. [1]

    Mechanisms of sound localization in mammals,

    B. Grothe, M. Pecka, and D. McAlpine, “Mechanisms of sound localization in mammals,” Physiological Reviews , vol. 90, no. 3, pp. 983–1012, 2010

    work page 2010

  4. [2]

    The generalized correlation method for estimation of time delay,

    C. Knapp and G. Carter, “The generalized correlation method for estimation of time delay,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 24, no. 4, pp. 320–327, 1976

    work page 1976

  5. [3]

    Use of the crosspower-spectrum phase in acoustic event location,

    M. Omologo and P. Svaizer, “Use of the crosspower-spectrum phase in acoustic event location,” IEEE Transactions on Speech and Audio Processing, vol. 5, no. 3, pp. 288–292, 1997

    work page 1997

  6. [4]

    Benesty, J

    J. Benesty, J. Cheng, and Y . Huang,Microphone Array Signal Processing , ser. Springer Topics in Signal Processing. Springer, 2008, vol. 1

    work page 2008

  7. [5]

    Multi-source tdoa estimation in reverberant audio using angular spectra and clustering,

    C. Blandin, A. Ozerov, and E. Vincent, “Multi-source tdoa estimation in reverberant audio using angular spectra and clustering,” Signal Processing, vol. 92, no. 8, pp. 1950–1960, 2012

    work page 1950

  8. [6]

    A learning-based approach to direction of arrival estimation in noisy and reverberant environments,

    X. Xiao, S. Zhao, X. Zhong, D. L. Jones, E. S. Chng, and H. Li, “A learning-based approach to direction of arrival estimation in noisy and reverberant environments,” in 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2015, pp. 2814–2818

    work page 2015

Show all 41 references
  1. [7]

    Steered response power for sound source localization: a tutorial review,

    E. Grinstein, E. Tengan, B. C ¸akmak, T. Dietzen, L. Nunes, T. Waterschoot, M. Brookes, and P. Naylor, “Steered response power for sound source localization: a tutorial review,” EURASIP Journal on Audio, Speech, and Music Processing, vol. 2024, 11 2024

    2024

  2. [8]

    Multiple emitter location and signal parameter estimation,

    R. Schmidt, “Multiple emitter location and signal parameter estimation,” IEEE Transactions on Antennas and Propagation , vol. 34, no. 3, pp. 276–280, 1986

    1986

  3. [9]

    Monaural sound localization,

    A. K. Fuchs, C. Feldbauer, and M. Stark, “Monaural sound localization,” in Proc. Interspeech, 2011, pp. 2521–2524

    2011

  4. [10]

    Exploiting deep neural networks and head movements for robust binaural localization of multiple sources in reverberant environments,

    N. Ma, T. May, and G. J. Brown, “Exploiting deep neural networks and head movements for robust binaural localization of multiple sources in reverberant environments,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 25, no. 12, pp. 2444–2453, 2017

    2017

  5. [11]

    Sound source localization using deep learning models,

    N. Yalta, K. Nakadai, and T. Ogata, “Sound source localization using deep learning models,” Journal of Robotics and Mechatronics , vol. 29, no. 1, pp. 37–48, 2017

    2017

  6. [12]

    Deep learning based multi-source localization with source splitting and its effectiveness in multi-talker speech recognition,

    A. S. Subramanian, C. Weng, S. Watanabe, M. Yu, and D. Yu, “Deep learning based multi-source localization with source splitting and its effectiveness in multi-talker speech recognition,” Computer Speech & Language, vol. 75, p. 101360, 2022

    2022

  7. [13]

    Self-supervised speech representation learning: A review,

    A. Mohamed, H.-y. Lee, L. Borgholt, J. D. Havtorn, J. Edin, C. Igel, K. Kirchhoff, S.-W. Li, K. Livescu, L. Maaløe, T. N. Sainath, and S. Watanabe, “Self-supervised speech representation learning: A review,” IEEE Journal of Selected Topics in Signal Processing , vol. 16, no. 6...

    2022

  8. [14]

    Sound localization based on phase difference enhancement using deep neural networks,

    J. Pak and J. W. Shin, “Sound localization based on phase difference enhancement using deep neural networks,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 27, no. 8, pp. 1335–1345, 2019

    2019

  9. [15]

    Estimation reliability function assisted sound source localization with enhanced steering vector phase difference,

    L. Cheng, X. Sun, D. Yao, J. Li, and Y . Yan, “Estimation reliability function assisted sound source localization with enhanced steering vector phase difference,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 29, pp. 421–435, 2021

    2021

  10. [17]

    Masked autoencoders that listen,

    P.-Y . Huang, H. Xu, J. Li, A. Baevski, M. Auli, W. Galuba, F. Metze, and C. Feichtenhofer, “Masked autoencoders that listen,” in Proc. NeurIPS, vol. 35, 2022, pp. 28 708–28 720

    2022

  11. [18]

    An unsupervised autore- gressive model for speech representation learning,

    Y . Chung, W. Hsu, H. Tang, and J. R. Glass, “An unsupervised autore- gressive model for speech representation learning,” in Proc. Interspeech, 2019, pp. 146–150

    2019

  12. [19]

    wav2vec 2.0: A framework for self-supervised learning of speech representations,

    A. Baevski, Y . Zhou, A. Mohamed, and M. Auli, “wav2vec 2.0: A framework for self-supervised learning of speech representations,” in Proc. NeurIPS, vol. 33, 2020, pp. 12 449–12 460

    2020

  13. [20]

    HuBERT: Self-supervised speech representation learning by masked prediction of hidden units,

    W.-N. Hsu, B. Bolte, Y .-H. H. Tsai, K. Lakhotia, R. Salakhutdinov, and A. rahman Mohamed, “HuBERT: Self-supervised speech representation learning by masked prediction of hidden units,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 29, pp. 3451–3460, 2021

    2021

  14. [21]

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

    S. Chen, C. Wang, Z. Chen, Y . Wu et al. , “WavLM: Large-scale self- supervised pre-training for full stack speech processing,” IEEE Journal of Selected Topics in Signal Processing , vol. 16, pp. 1505–1518, 2021

    2021

  15. [22]

    Beats: audio pre-training with acoustic tokenizers,

    S. Chen, Y . Wu, C. Wang, S. Liu, D. Tompkins, Z. Chen, W. Che, X. Yu, and F. Wei, “Beats: audio pre-training with acoustic tokenizers,” in Proceedings of the 40th International Conference on Machine Learning . JMLR.org, 2023

    2023

  16. [23]

    Representation learning with contrastive predictive coding,

    A. van den Oord, Y . Li, and O. Vinyals, “Representation learning with contrastive predictive coding,” ArXiv, vol. abs/1807.03748, 2018

    2018 arXiv

  17. [24]

    Towards robust speech representation learning for thousands of languages,

    W. Chen, W. Zhang, Y . Peng, X. Li, J. Tian, J. Shi, X. Chang, S. Maiti, K. Livescu, and S. Watanabe, “Towards robust speech representation learning for thousands of languages,” in Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing . Associa...

    2024

  18. [25]

    A noise-robust self-supervised pre-training model based speech representation learning for automatic speech recognition,

    Q.-S. Zhu, J. Zhang, Z.-Q. Zhang, M.-H. Wu, X. Fang, and L.-R. Dai, “A noise-robust self-supervised pre-training model based speech representation learning for automatic speech recognition,” in Proc. ICASSP, 2022, pp. 3174–3178

    2022

  19. [26]

    Joint separation and localization of moving sound sources based on neural full-rank spatial covariance analysis,

    H. Munakata, Y . Bando, R. Takeda, K. Komatani, and M. Onishi, “Joint separation and localization of moving sound sources based on neural full-rank spatial covariance analysis,” IEEE Signal Processing Letters , vol. 30, pp. 384–388, 2023

    2023

  20. [27]

    Unssor: Unsupervised neural speech separation by leveraging over-determined training mixtures,

    Z.-Q. Wang and S. Watanabe, “Unssor: Unsupervised neural speech separation by leveraging over-determined training mixtures,” in Proc. NeurIPS, vol. 36, 2023, pp. 34 021–34 042

    2023

  21. [29]

    Librispeech: An ASR corpus based on public domain audio books,

    V . Panayotov, G. Chen, D. Povey, and S. Khudanpur, “Librispeech: An ASR corpus based on public domain audio books,” in Proc. ICASSP , 2015, pp. 5206–5210

    2015

  22. [30]

    Libri-Light: A benchmark for asr with limited or no supervision,

    J. Kahn, M. Rivi `ere, W. Zheng, E. Kharitonov, Q. Xu, P. E. Mazar ´e, J. Karadayi, V . Liptchinsky, R. Collobert, C. Fuegen, T. Likhomanenko, G. Synnaeve, A. Joulin, A. Mohamed, and E. Dupoux, “Libri-Light: A benchmark for asr with limited or no supervision,” in Proc. ICASSP ...

    2020

  23. [31]

    HRTF-DATABASE,

    Institut f ¨ur Schallforschung der ¨Osterreichischen Akademie der Wissenschaften, “HRTF-DATABASE,” https://www.oeaw.ac.at/isf/das- institut/software/hrtf-database, 2024

    2024

  24. [32]

    Signal-informed dnn-based doa estimation combining an external microphone and gcc-phat features,

    U. Kowalk, S. Doclo, and J. Bitzer, “Signal-informed dnn-based doa estimation combining an external microphone and gcc-phat features,” in Proc. IWAENC, 2022, pp. 1–5

    2022

  25. [33]

    Geometry-aware doa estimation using a deep neural network with mixed-data input features,

    ——, “Geometry-aware doa estimation using a deep neural network with mixed-data input features,” in Proc. ICASSP, 2023, pp. 1–5

    2023

  26. [34]

    Deep learning-based speech specific source localization by using binaural and monaural microphone arrays in hearing aids,

    P. Goli and S. van de Par, “Deep learning-based speech specific source localization by using binaural and monaural microphone arrays in hearing aids,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 31, pp. 1652–1666, 2023

    2023

  27. [35]

    Regression and clas- sification for direction-of-arrival estimation with convolutional recurrent neural networks,

    Z. Tang, J. D. Kanu, K. Hogan, and D. Manocha, “Regression and clas- sification for direction-of-arrival estimation with convolutional recurrent neural networks,” in Interspeech, 2019

    2019

  28. [36]

    Conformer: Convolution- augmented transformer for speech recognition,

    A. Gulati, J. Qin, C.-C. Chiu, N. Parmar, Y . Zhang, J. Yu, W. Han, S. Wang, Z. Zhang, Y . Wu, and R. Pang, “Conformer: Convolution- augmented transformer for speech recognition,” in Proc. Interspeech , Shanghai, China, 10 2020, pp. 5036–5040

    2020

  29. [37]

    The 1st clarity prediction challenge: A machine learning challenge for hearing aid intelligibility prediction

    J. Barker, M. Akeroyd, T. J. Cox, J. F. Culling, J. Firth, S. Graetzer, H. Griffiths, L. Harris, G. Naylor, Z. Podwinska et al. , “The 1st clarity prediction challenge: A machine learning challenge for hearing aid intelligibility prediction.” in Proc. Interspeech, 2022, pp. 3508–3512

    2022

  30. [38]

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

    T. Ko, V . Peddinti, D. Povey, M. L. Seltzer, and S. Khudanpur, “A study on data augmentation of reverberant speech for robust speech recognition,” in Proc. ICASSP, 2017, pp. 5220–5224

    2017

  31. [39]

    anf-generator,

    International Audio Laboratories Erlangen, “anf-generator,” https://github. com/audiolabs/anf-generator, 2025

    2025

  32. [40]

    Speech enhancement using long short-term memory based recurrent neural networks for noise robust speaker verification,

    M. Kolbœk, Z.-H. Tan, and J. Jensen, “Speech enhancement using long short-term memory based recurrent neural networks for noise robust speaker verification,” in 2016 IEEE Spoken Language Technology Workshop (SLT), 2016, pp. 305–311

    2016

  33. [41]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” in International Conference on Learning Representations , 2019

    2019

Pith tools

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