Pith. sign in

REVIEW 4 major objections 5 minor 38 references

Zero-Shot KWS for Children's Speech using Layer-Wise Features from SSL Models

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 Features from layer 22 of Wav2Vec2 let a keyword spotter trained on adult speech work on children's speech, reaching an ATWV of 0.691.

desk verdict Useful zero-shot child-KWS idea, but test-set layer selection and impossible ATWV values make the headline numbers unreliable. read the letter →

arxiv 2508.21248 v1 pith:OZDQKO7Y submitted 2025-08-28 eess.AS cs.AIcs.HCcs.SDeess.SP

Subham Kutum , Abhijit Sinha , Hemant Kumar Kathania , Sudarsana Reddy Kadiri , Mahesh Chandra Govil This is my paper · ORCID
classification eess.AScs.AIcs.HCcs.SDeess.SP
keywords zero-shotkeywordspottingchildren'sspeechself-supervisedlearningWav2Vec2HuBERTData2Veclayer-wisefeaturesnoiserobustness
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 show that a keyword-spotting (KWS) system can work on children's speech without ever training on a child's voice. The trick is to replace traditional MFCC features with frame-level embeddings taken from individual hidden layers of pretrained self-supervised speech models—Wav2Vec2, HuBERT, and Data2Vec—and feed those into a standard DNN-HMM KWS pipeline trained on adult speech. On the PFSTAR children's test set, the best model (Wav2Vec2, layer 22) reaches ATWV 0.691 for 30 keywords, far above the MFCC baseline. The claim matters because it points to a privacy-preserving path: children's KWS without collecting or storing child speech recordings. The same layer also improves robustness to noise and transfers to another children's dataset, CMU Kids.

What carries the argument

Layer-wise SSL embeddings: 25 hidden layers, each outputting 1024-dimensional frame-level vectors from CNN-plus-Transformer speech encoders. The KWS machinery is a DNN-HMM acoustic model with lattice-based index-and-search keyword scoring. The mechanism that carries the argument is layer choice: low layers give negative scores, middle layers start to work, and late layers—particularly Wav2Vec2 layer 22—supply the abstract subword- and word-level features that survive the acoustic mismatch of children's speech.

What would settle it

Split the PFSTAR children's data into a separate layer-selection set and a held-out test set; pick the best layer on the selection set, then measure ATWV on the held-out set. If the SSL-layer advantage over MFCC shrinks to near zero or disappears, the central claim of layer-22 superiority would be falsified.

Watch

Extended reading notes

Core claim

The central discovery is that the deepest transformer layers of Wav2Vec2 carry representations that transfer across the adult-to-child domain shift. Among the 25 hidden layers examined, layer 22 consistently outperforms both early spectral layers and the middle layers, and it beats the MFCC baseline by a wide margin on every keyword set. For 30 keywords the system reports ATWV 0.691, MTWV 0.700, false-alarm probability 0.016 and miss probability 0.055; it also beats prior children's-KWS methods on the same PFSTAR test set. The authors attribute the gain to Wav2Vec2's contrastively trained, noise-invariant, context-rich upper-layer embeddings, which preserve the phonetic detail needed for key

Load-bearing premise

The paper chooses layer 22 by looking at the test-set results and then reports that layer's score as its result, so the headline numbers assume the test set can double as a model-selection set.

Editorial extensions

If this is right

  • If this result is right, KWS for children no longer requires labeled child speech, lowering privacy and data-collection barriers.
  • A single late SSL layer can replace MFCC features in a standard KWS pipeline with no architectural change beyond feature extraction.
  • The age-group breakdown implies the same system is usable for older children but would need adaptation for younger children aged 4 to 6.
  • Because the best layer stays layer 22 across keyword sets and datasets, the paper suggests layer choice is a stable property of the model rather than of the keyword list.
  • Noise experiments show SSL features degrade far more gracefully than MFCCs, so the same approach could also help adult KWS in noisy environments.

Reading between the lines

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

  • Beyond the paper: since the best layer was selected on the same test set it evaluates, the absolute ATWV numbers are likely optimistic; a held-out validation layer choice would be needed to know the true zero-shot gap.
  • Beyond the paper: the paper tests additive noise but not reverberation or overlapping speech; those are the next plausible stress tests for the claimed robustness.
  • Beyond the paper: fusing features from several late layers (for example, layers 20-24) might improve on single-layer 22, an experiment the paper does not run.
  • Beyond the paper: the CMU generalization used a different adult training corpus for American English, so the framework may be accent-flexible, but the cross-corpus comparison is not apples-to-apples.
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 a zero-shot keyword spotting system for children's speech. It extracts layer-wise features from three self-supervised speech models (Wav2Vec2, HuBERT, Data2Vec), trains a Kaldi-based DNN KWS system on adult speech (WSJCAM0), and evaluates on children's speech (PFSTAR). The authors report that Wav2Vec2 layer 22 performs best, with ATWV 0.691, MTWV 0.7003, Pfa 0.0164, Pmiss 0.0547 on a 30-keyword set. They also present experiments across keyword-set sizes, age groups, noise conditions, and a second children's corpus (CMU Kids), and compare against prior PFSTAR-based KWS studies. The central claim is that SSL features, especially from a late Wav2Vec2 layer, improve zero-shot KWS for children's speech over a traditional MFCC baseline.

Significance. If the results were obtained from a sound evaluation protocol, the paper would make a useful contribution: a systematic layer-wise comparison of SSL features for zero-shot children's KWS, with coverage of age effects, noise robustness, and cross-corpus generalization, is valuable for a relatively underexplored application. The manuscript's strengths are its breadth of experiments and the clear practical framing. However, the headline result is compromised by the method used to select the 'best' SSL layer: every layer is evaluated on the PFSTAR test set and the peak is then reported as the system's performance. This introduces optimistic bias into all primary quantitative claims, including the state-of-the-art comparison. In addition, several reported ATWV values exceed the theoretical maximum of 1.0, and the claimed statistical significance is not supported by any reported test statistics. These issues are load-bearing for the paper's central conclusions.

major comments (4)
  1. [§4.1, Figs. 2–3, Table 2] The best-performing SSL layer is selected using the PFSTAR test set itself. Section 3.1 states that PFSTAR is used 'for testing', and Figs. 2–3 plot ATWV/MTWV for all 25 layers on that test set. Table 2 then reports the best layer (e.g., layer 22 for Wav2Vec2) as the final result. This is test-set selection: the reported ATWV of 0.691 is the maximum over 25 candidate configurations evaluated on the same data used to measure final performance. The resulting number is optimistically biased, and the later age, noise, CMU, and comparison experiments all inherit this test-selected layer. The zero-shot evaluation protocol does not establish that layer 22 would be chosen without access to the test labels. A held-out validation set, nested cross-validation, or a pre-specified layer-selection rule is required before the headline numbers can be taken as unbiased estimates.
  2. [Tables 8–9] Several reported ATWV scores are impossible. ATWV is a term-weighted value that is bounded above by 1 (the paper itself states that TWV should be closer to 1 for good performance). Table 9 reports ATWV = 5.861 for Layer 22 under 'Crowd' at 5 dB for 10 keywords, ATWV = 12.725 for the MFCC baseline under 'Birds Chirping' at 5 dB for 10 keywords, and ATWV = 3.551 for Layer 22 under 'Birds Chirping' at 10 dB for 20 keywords. These values exceed 1.0 by large margins, indicating a scoring or tabulation error. Since the noise-robustness claims rely on these tables, the numerical reliability of the entire noise analysis is in question.
  3. [§4.1 and Abstract] The manuscript claims that 'statistical analyses including paired t-tests and Wilcoxon signed-rank tests' confirm that the observed improvements are statistically significant, but no test statistics, p-values, sample sizes, or descriptions of the paired structure are reported anywhere. The claim cannot be verified. Either the full test results should be reported, or the significance claim should be removed from the abstract and Section 4.1.
  4. [§4.6, Table 10] The comparison with prior work is unfair because the proposed method's layer index was chosen by searching over the PFSTAR test set, while the prior methods (Pattanayak et al. 2022, 2021; Shahnawazuddin et al. 2019) did not receive equivalent test-set-based model selection. The reported margin over these baselines therefore reflects not only the benefit of SSL features but also the benefit of test-set tuning. To support 'state-of-the-art across all keyword sets', the authors must compare under a common model-selection protocol, e.g., by fixing the layer on a held-out development set before evaluating on the test set.
minor comments (5)
  1. [Abstract] Typo: 'gerneralize' should be 'generalize'.
  2. [§4.1] The text states 'Wav2Vec2 reaches its best at layer 22 (ATWV 0.6917, MTWV 0.7003)' without specifying the keyword set; Table 2 shows ATWV 0.535 for 10 keywords, 0.661 for 20, and 0.691 for 30. Please clarify which keyword set is being referenced.
  3. [Table 3] The baseline (no modification) rows list identical ATWV/MTWV values in both the 'Modified PFSTAR' and 'Augmented WSJCAM0' columns. This is confusing; please clarify what the columns represent and whether the baseline is the same in both settings.
  4. [References] Reference [26] contains a typo: 'Doddingtion' should be 'Doddington'.
  5. [§3.2] The DNN acoustic model is not fully specified: no architecture details (number of hidden layers, units, activation, training epochs) are given. Since the paper is about layer-wise SSL features, the downstream model's capacity is relevant for reproducibility.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline ATWV 0.691 is the peak of a 25-layer search on the PFSTAR test set; layer 22 is 'best' by construction because no held-out validation is used.

  1. fitted input called prediction [Section 3.1 (Database); Section 4.1 and Table 2 (SSL Layer-wise Features Performance)]
    "The best-performing layers for each model and keyword set are summarized in Table 2. A subset of this dataset, containing 1.1 hours of audio from 60 speakers, including 28 females, was used for testing."

    The paper evaluates all 25 SSL layers on the PFSTAR test set (Figs. 2-3), then selects layer 22 because it has the highest ATWV/MTWV and reports that same test-set value (ATWV 0.691, MTWV 0.700) as the final system performance. Section 3.1 describes PFSTAR only as 'testing'; no development/validation set is used to choose the layer. Thus the headline number is the argmax over 25 configurations evaluated on the exact data used for final scoring. 'Layer 22 is best' is true by construction: it is the maximum of the search. The reported ATWV is the fitted peak, not an unbiased estimate for a pre-specified system, and the state-of-the-art comparison in Table 10 is therefore against systems not given this per-keyword-set test-set selection. Downstream age, noise, and CMU experiments reuse this t

full rationale

The paper's central contribution is an empirical measurement rather than a derivation: SSL layer features from Wav2Vec2/HuBERT/Data2Vec are fed to a Kaldi DNN-KWS pipeline and evaluated on child speech. There is no equation-level derivation whose conclusion is equivalent to its premise, and the citations to prior work (including overlapping-author references [17], [31]-[33]) are used only for standard preprocessing/augmentation choices; they are not load-bearing for the main result. The significant circularity concern is the layer-selection protocol: the system's best layer is chosen by evaluating every layer on the PFSTAR test set, and the chosen layer's test-set score is then reported as the achieved performance. This is a fitted-input-called-prediction pattern: the reported ATWV/MTWV values are maxima over the evaluation data, making the 'optimal layer 22' claim self-fulfilling and the performance estimate optimistically biased. No held-out validation set is described. Numerical anomalies in Tables 8-9 (ATWV > 1, e.g., 5.861 and 12.725) may indicate scoring errors but are not circularity. Overall, the central claim reduces in part to the test-set fit, so the score is 6 rather than 0-2, though the general claim that SSL features can outperform MFCCs for child KWS retains independent empirical content.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the choice of the best SSL layer and model, which is made on the test set rather than by a principled criterion. No new entities are introduced. The SSL feature extractors, Kaldi recipe, and datasets are external inputs.

free parameters (1)
  • selected SSL layer index (layer 22 for Wav2Vec2) = 22
    Chosen as the layer with maximum ATWV on the PFSTAR test set (Table 2, Section 4.1). This selection is done on the test data, so all reported headline numbers are conditional on this test-set-tuned choice.
assumptions (4)
  • domain assumption Kaldi DNN acoustic model trained on adult WSJCAM0 speech is a valid acoustic model for keyword search in a zero-shot children's setting.
    Section 3.2, the DNN-HMM is used to generate lattices for the PFSTAR test set without domain adaptation.
  • domain assumption Pretrained SSL models (Wav2Vec2, HuBERT, Data2Vec) are used as fixed feature extractors and their features are sufficiently invariant across the adult-child domain gap.
    Sections 2 and 3.3; no fine-tuning or domain adaptation is applied.
  • domain assumption CMVN applied to SSL features normalizes the distribution mismatch enough for the Kaldi DNN to work on child speech.
    Section 3.3, final paragraph.
  • standard math Standard Kaldi KWS pipeline's lattice-based indexing and scoring is assumed correct.
    Section 3.2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Zero-Shot KWS for Children's Speech using Layer-Wise Features from SSL Models." pith.science (2026). https://pith.science/paper/OZDQKO7Y

@misc{pith2026250821248,
  author       = {Pith},
  title        = {Pith review of: Zero-Shot KWS for Children's Speech using Layer-Wise Features from SSL Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OZDQKO7Y}},
  note         = {Machine review of arXiv:2508.21248}
}
read the original abstract

Numerous methods have been proposed to enhance Keyword Spotting (KWS) in adult speech, but children's speech presents unique challenges for KWS systems due to its distinct acoustic and linguistic characteristics. This paper introduces a zero-shot KWS approach that leverages state-of-the-art self-supervised learning (SSL) models, including Wav2Vec2, HuBERT and Data2Vec. Features are extracted layer-wise from these SSL models and used to train a Kaldi-based DNN KWS system. The WSJCAM0 adult speech dataset was used for training, while the PFSTAR children's speech dataset was used for testing, demonstrating the zero-shot capability of our method. Our approach achieved state-of-the-art results across all keyword sets for children's speech. Notably, the Wav2Vec2 model, particularly layer 22, performed the best, delivering an ATWV score of 0.691, a MTWV score of 0.7003 and probability of false alarm and probability of miss of 0.0164 and 0.0547 respectively, for a set of 30 keywords. Furthermore, age-specific performance evaluation confirmed the system's effectiveness across different age groups of children. To assess the system's robustness against noise, additional experiments were conducted using the best-performing layer of the best-performing Wav2Vec2 model. The results demonstrated a significant improvement over traditional MFCC-based baseline, emphasizing the potential of SSL embeddings even in noisy conditions. To further generalize the KWS framework, the experiments were repeated for an additional CMU dataset. Overall the results highlight the significant contribution of SSL features in enhancing Zero-Shot KWS performance for children's speech, effectively addressing the challenges associated with the distinct characteristics of child speakers.

Figures

Figures reproduced from arXiv: 2508.21248 by the authors.

Figure 1
Figure 1. Block diagram illustrating the proposed keyword spotting (KWS) [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. ATWV and MTWV results for 10, 20 and 30 keywords using layer-wise features from SSL models: Wav2Vec2, HuBERT and Data2Vec, on the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Pfa and Pmiss results for 10, 20 and 30 keywords using layer-wise features from SSL models: Wav2Vec2, HuBERT and Data2Vec, on the PFSTAR dataset. by NIST for spoken term detection, including Actual Term Weighted Value (ATWV), Maximum Term Weighted Value (MTWV), probability of false alarm (Pfa) and probability of miss (Pmiss) [26]. This approach aligns with prior studies [27, 28]. For a good KWS performance, TWV valu… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 35 canonical work pages

  1. [1]

    Weintraub, Lvcsr log-likelihood ratio scoring for keyword spotting, in: International Conference on Acoustics, Speech, and Signal Processing, V ol

    M. Weintraub, Lvcsr log-likelihood ratio scoring for keyword spotting, in: International Conference on Acoustics, Speech, and Signal Processing, V ol. 1, IEEE, 1995, pp. 297–300

    work page 1995

  2. [2]

    Sz ¨oke, P

    I. Sz ¨oke, P. Schwarz, P. Matejka, L. Burget, M. Karafi´at, M. Fapso, J. Cer- nock`y, Comparison of keyword spotting approaches for informal contin- uous speech., in: Interspeech, 2005, pp. 633–636

    work page 2005

  3. [3]

    R. C. Rose, D. B. Paul, A hidden markov model based keyword recogni- tion system, in: International conference on acoustics, speech, and signal processing, IEEE, 1990, pp. 129–132

    work page 1990

  4. [4]

    D. Can, M. Saraclar, Lattice indexing for spoken term detection, IEEE Transactions on Audio, Speech, and Language Processing 19 (8) (2011) 2338–2347

    work page 2011

  5. [5]

    Online Keyword Spotting with a Character-Level Recurrent Neural Network

    K. Hwang, M. Lee, W. Sung, Online keyword spotting with a character- level recurrent neural network, ArXiv abs/1512.08903 (2015). URL https://api.semanticscholar.org/CorpusID:6680389

    work page Pith review arXiv 2015

  6. [6]

    L ´opez-Espejo, Z.-H

    I. L ´opez-Espejo, Z.-H. Tan, J. H. Hansen, J. Jensen, Deep spoken key- word spotting: An overview, IEEE Access 10 (2021) 4169–4199

    work page 2021

  7. [7]

    Mazumder, C

    M. Mazumder, C. R. Banbury, J. Meyer, P. Warden, V . J. Reddi, Few-shot keyword spotting in any language, in: Interspeech, 2021. URL https://api.semanticscholar.org/CorpusID:233025251

    work page 2021

  8. [8]

    A. Berg, M. O’Connor, M. T. Cruz, Keyword transformer: A self- attention model for keyword spotting, in: Interspeech 2021, 2021, pp. 4249–4253. doi:10.21437/Interspeech.2021-1286

    work page doi:10.21437/interspeech.2021-1286 2021

Show all 38 references
  1. [9]

    Warden, Speech commands: A dataset for limited-vocabulary speech recognition, arXiv preprint arXiv:1804.03209 (2018)

    P. Warden, Speech commands: A dataset for limited-vocabulary speech recognition, arXiv preprint arXiv:1804.03209 (2018)

    2018 arXiv

  2. [10]

    Hussain, V

    S. Hussain, V . Nguyen, S. Zhang, E. Visser, Multi-task voice activated framework using self-supervised learning, in: IEEE International Confer- ence on Acoustics, Speech and Signal Processing (ICASSP), 2022, pp. 6137–6141

    2022

  3. [11]

    J. Mørk, H. S. Bovbjerg, G. Kiss, Z.-H. Tan, Noise-robust key- word spotting through self-supervised pretraining, arXiv preprint arXiv:2403.18560 (2024)

    2024 arXiv

  4. [12]

    Baevski, Y

    A. Baevski, Y . Zhou, A. Mohamed, M. Auli, wav2vec 2.0: A framework for self-supervised learning of speech representations, Advances in neural information processing systems 33 (2020) 12449–12460

    2020

  5. [13]

    S. Chen, C. Wang, Z. Chen, Y . Wu, S. Liu, Z. Chen, J. Li, N. Kanda, T. Yoshioka, X. Xiao, et al., Wavlm: Large-scale self-supervised pre- training for full stack speech processing, IEEE Journal of Selected Topics in Signal Processing 16 (6) (2022) 1505–1518

    2022

  6. [14]

    P. Zhu, J. W. Bartel, D. Agarwal, K. Partridge, H. J. Park, Q. Wang, Ge2e- kws: Generalized end-to-end training and evaluation for zero-shot key- word spotting, in: 2024 IEEE Spoken Language Technology Workshop (SLT), IEEE, 2024, pp. 999–1006

    2024

  7. [15]

    Huang, M

    R. Huang, M. Wiesner, L. P. Garcia-Perera, D. Povey, J. Trmal, S. Khu- danpur, Building keyword search system from end-to-end asr systems, in: ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2023, pp. 1–5

    2023

  8. [16]

    Radford, J

    A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, I. Sutskever, Robust speech recognition via large-scale weak supervision, in: Interna- tional conference on machine learning, PMLR, 2023, pp. 28492–28518

    2023

  9. [17]

    Sinha, M

    A. Sinha, M. Singh, S. R. Kadiri, M. Kurimo, H. K. Kathania, E ffect of speech modification on wav2vec2 models for children speech recognition, in: International Conference on Signal Processing and Communications (SPCOM), IEEE, 2024, pp. 1–5

    2024

  10. [18]

    J. Li, M. A. Hasegawa-Johnson, N. L. McElwain, Analysis of self- supervised speech models on children’s speech and infant vocalizations, IEEE International Conference on Acoustics, Speech, and Signal Process- ing Workshops (ICASSPW) (2024) 550–554. URL https://api.semanticscho...

    2024

  11. [19]

    M. Yu, X. Jin, B. Wan, G. Wang, A few-shot speech keyword spot- ting method based on self-supervise learning, in: 2023 16th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI), IEEE, 2023, pp. 1–5

    2023

  12. [20]

    Jacobs, Multilingual acoustic word embeddings for zero-resource lan- guages, arXiv preprint arXiv:2401.10543 (2024)

    C. Jacobs, Multilingual acoustic word embeddings for zero-resource lan- guages, arXiv preprint arXiv:2401.10543 (2024)

    2024 arXiv

  13. [21]

    W.-N. Hsu, B. Bolte, Y .-H. H. Tsai, K. Lakhotia, R. Salakhutdinov, A. Mohamed, Hubert: Self-supervised speech representation learning by masked prediction of hidden units, IEEE /ACM Transactions on Audio, Speech, and Language Processing

  14. [22]

    Baevski, W.-N

    A. Baevski, W.-N. Hsu, Q. Xu, A. Babu, J. Gu, M. Auli, Data2vec: A general framework for self-supervised learning in speech, vision and lan- guage, in: International conference on machine learning, PMLR, 2022, pp. 1298–1312

    2022

  15. [23]

    Povey, A

    D. Povey, A. Ghoshal, G. Boulianne, L. Burget, O. Glembek, N. Goel, M. Hannemann, P. Motlicek, Y . Qian, P. Schwarz, et al., The kaldi speech recognition toolkit, in: IEEE 2011 workshop on automatic speech recog- nition and understanding, IEEE Signal Processing Society

    2011

  16. [24]

    Robinson, J

    T. Robinson, J. Fransen, D. Pye, J. Foote, S. Renals, Wsjcamo: a british english speech corpus for large vocabulary continuous speech recogni- tion, in: 1995 International Conference on Acoustics, Speech, and Signal Processing, V ol. 1, IEEE, pp. 81–84

    1995

  17. [25]

    Batliner, M

    A. Batliner, M. Blomberg, S. D’Arcy, D. Elenius, D. Giuliani, M. Gerosa, C. Hacker, M. Russell, S. Steidl, M. Wong, The pf star children’s speech corpus (2005)

    2005

  18. [26]

    J. G. Fiscus, J. Ajot, J. S. Garofolo, G. Doddingtion, Results of the 2006 spoken term detection evaluation, in: Proc. sigir, V ol. 7, 2007, pp. 51–57

    2006

  19. [27]

    Pattanayak, G

    B. Pattanayak, G. Pradhan, Significance of single frequency filter for the development of children’s kws system., in: INTERSPEECH, 2022, pp. 3183–3187

    2022

  20. [28]

    Shahnawazuddin, K

    S. Shahnawazuddin, K. Maity, G. Pradhan, Improving the performance of keyword spotting system for children’s speech through prosody modifica- tion, Digital Signal Processing 86 (2019) 11–18

    2019

  21. [29]

    Wegmann, A

    S. Wegmann, A. Faria, A. Janin, K. Riedhammer, N. Morgan, The tao of atwv: Probing the mysteries of keyword search performance, in: 2013 IEEE Workshop on Automatic Speech Recognition and Understanding, IEEE, pp. 192–197

    2013

  22. [30]

    X. Zhu, G. T. Beauregard, L. L. Wyse, Real-time signal estimation from modified short-time fourier transform magnitude spectra, IEEE Transac- tions on Audio, Speech, and Language Processing (July 2007)

    2007

  23. [31]

    H. K. Kathania, W. Ahmad, S. Shahnawazuddin, A. B. Samaddar, Ex- plicit pitch mapping for improved children’s speech recognition, Circuits, Systems, and Signal Processing (2018)

    2018

  24. [32]

    Kathania, M

    H. Kathania, M. Singh, T. Gr ´osz, M. Kurimo, Data augmentation us- ing prosody and false starts to recognize non-native children’s speech, in: Proc. Interspeech 2020, 2020

    2020

  25. [33]

    H. K. Kathania, S. R. Kadiri, P. Alku, M. Kurimo, Study of formant modi- fication for children asr, in: ICASSP 2020-2020 IEEE International Con- ference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2020, pp. 7429–7433

    2020

  26. [34]

    Eskenazi, J

    M. Eskenazi, J. Mostow, D. Gra ff, The cmu kids corpus, Linguistic Data Consortium 11 (1997)

    1997

  27. [35]

    Panayotov, G

    V . Panayotov, G. Chen, D. Povey, S. Khudanpur, Librispeech: An asr corpus based on public domain audio books, in: 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2015, pp. 5206–5210. doi:10.1109/ICASSP.2015.7178964

    2015

  28. [36]

    Varga, H

    A. Varga, H. J. Steeneken, Assessment for automatic speech recognition: Ii. noisex-92: A database and an experiment to study the e ffect of addi- tive noise on speech recognition systems, Speech communication 12 (3) (1993) 247–251

    1993

  29. [37]

    Snyder, G

    D. Snyder, G. Chen, D. Povey, Musan: A music, speech, and noise corpus, arXiv preprint arXiv:1510.08484 (2015)

    2015 arXiv

  30. [38]

    Pattanayak, G

    B. Pattanayak, G. Pradhan, Pitch-robust acoustic feature using single frequency filtering for children’s kws, Pattern Recognition Letters 150 (2021) 183–188

    2021

Pith tools

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