Making deep neural networks work for medical audio: representation, compression and domain adaptation

Charles C Onu

REVIEW 4 major objections 6 minor 263 references

Making deep neural networks work for medical audio: representation, compression and domain adaptation

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This thesis claims a newborn's cry, recorded on a smartphone, can screen for neurological injury with 92.5% AUC once deep learning is adapted for medical audio.

desk verdict The engineering contributions are real and some are already peer-reviewed, but the thesis's headline 92.5% AUC for neuro-injury detection is not traceable to its own experiments and the dataset counts don't reconcile, so the clinical claim is unverified until fixed. read the letter →

arxiv 2506.13970 v1 pith:XLFXM6I6 submitted 2025-05-24 cs.SD cs.AIeess.AS

Charles C Onu This is my paper · ORCID

classification cs.SDcs.AIeess.AS

keywords infantcryanalysismedicalaudiotransferlearningself-supervisedtensordecompositionmodelcompressiondomainadaptationneurologicalinjuryscreening

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 thesis tries to establish that infant cry analysis can become a working clinical screening tool for neurological injury, provided the machine-learning pipeline is rebuilt for the realities of medical audio: scarce labels, noisy recordings, and hospitals that record differently. The author brings four techniques to bear—transfer learning from adult speech, self-supervised pre-training, end-to-end model compression with tensor decomposition, and domain adaptation—and reports that the assembled system detects neurological injury from newborns' cries with an AUC of 92.5% (88.7% sensitivity at 80% specificity) across five hospitals on three continents. If those numbers hold, a 30-second cry recording taken on a smartphone at birth could screen for brain injury from asphyxia in settings where specialists are unavailable.

What carries the argument

The clinical claim is carried by a three-stage training pipeline: a CNN-style encoder (CNN14) is first pre-trained with the SimCLR contrastive objective on a large general-audio collection, then re-adapted with SimCLR on unlabeled cry recordings while replaying a portion of the original audio to prevent forgetting, and finally fine-tuned with clinical labels. The compression strand rests on a named identity: in a fully tensorized recurrent cell, the gate weight matrices are row-wise concatenated and encoded as a single tensor-train, so every gate matrix is a linear mixture $W_i = \sum_{\alpha} V_{i,\alpha} M_{\alpha}$ of $r_0$ shared tensor-train matrices—an arrangement that multiplies the usual tensor-train parameter reduction by the gate count and supplies implicit regularization.

What would settle it

A strict leave-one-hospital-out evaluation would settle the claim: train on four hospitals, test on the fifth, repeat for all five, and check whether AUC holds up when the test hospital's site signature and label style are absent from training. A second check is re-adjudicating a random sample of Sarnat labels with a blinded central panel and re-measuring AUC; if out-of-hospital performance falls to chance, or the model turns out to be predicting recording conditions rather than injury, the 92.5% figure is not a diagnostic signal.

Watch

Extended reading notes

Core claim

The central claim is that the newborn cry can function as a vital sign: the acoustic content of a 30-second cry recorded on a smartphone soon after birth carries enough information about neurological state that a deep model can flag injury caused by birth asphyxia. On a database built across five hospitals and three continents, the assembled system detects neurological injury with an AUC of 92.5% (88.7% sensitivity at 80% specificity). The route to that result is a chain of technical claims: representations from large adult-speech corpora transfer to cries and handle noise better than training from scratch; a three-stage self-supervised pipeline (contrastive pre-training on general audio, contrastive adaptation on unlabeled cries, then supervised fine-tuning) delivers the top accuracy and keeps roughly 70% AUC with only 3% of labeled data; and domain adaptation, notably a target-noise-injection method that needs no target labels, lifts cross-hospital target accuracy by 7.2% without hurting source accuracy. The same body of work contributes the CryCeleb public cry dataset and shows that fully tensorized recurrent networks cut parameters 50 to 653 times with equal or better accuracy, so the screening models are small enough to deploy.

Load-bearing premise

The load-bearing premise is that the clinical labels are true: each infant's neurological injury is graded by a clinician with a modified Sarnat score at each hospital, and if those grades are applied inconsistently across the five sites or are entangled with how and where the recording was made, the 92.5% AUC would not measure the cry's real diagnostic value.

Editorial extensions

If this is right

A cry recorded at birth predicts neurological injury with 92.5% AUC (88.7% sensitivity at 80% specificity) in the five-hospital study; the headline consequence is a low-cost, contact-free screening test for asphyxia-related brain injury.
The self-supervised cry-adaptation stage cuts labeled-data needs dramatically: about 70% AUC with only 3% of labeled data, and better than trained-from-scratch baselines at 20%, so smaller clinical studies can build working models.
Fully tensorized RNNs reach 50 to 653 times parameter compression while maintaining or improving accuracy (speaker-verification equal error rate drops from 7.33% to 4.34% on LibriSpeech), putting cry-analysis models in reach of smartphones and edge devices.
Domain adaptation transfers models across hospitals: the best adapted model gains 7.2% target-domain accuracy without losing source accuracy, addressing the dataset-bias barrier to multi-site deployment.
Adult speech is a workable source domain for cry models: word-recognition pre-training outperforms the SVM baseline for asphyxia detection (86.5% vs 84.4% unweighted average recall) and the neural models stay more stable under noise and signal loss.

Reading between the lines

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

A fair test of the 92.5% AUC claim would be strict leave-one-hospital-out evaluation: if performance collapses on unseen hospitals once site-specific acoustic signatures (which the paper's own 'name-the-hospital' experiment shows exist) are removed, part of the reported accuracy may reflect recording-site artifacts rather than the cry's medical signal.
Because the labels are modified Sarnat scores assigned by different clinicians at five sites, label noise is the load-bearing risk; re-adjudicating a subset with a blinded central panel and re-measuring AUC would quantify how much of the figure is diagnostic signal versus annotation style.
The gate-concatenation identity suggests a general recipe for gated architectures—jointly tensorizing all gates of a cell gives compression proportional to gate count—so extending the same treatment to attention-based audio models is a natural next step that the thesis itself lists as future work.
The CryCeleb result that cry-based identity verification stays far behind adult speaker verification (equal error rate near 28% after fine-tuning) implies cries vary widely within an infant; a practical corollary is that screening systems should pool evidence across multiple cry episodes rather than judge single expirations.

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

Here's my take. The methods chapters are mostly solid and several rest on prior peer-reviewed papers. The fully tensorized RNN with gate concatenation is a genuine extension of earlier tensorized RNN work, and the reported compression ratios (50-650x) with maintained or improved accuracy on speaker verification are credible; the open-source code helps. CryCeleb is a real public dataset with over six hours of recordings and a documented baseline, which is useful for the community. The SSL cry-adaptation pipeline with replay is sensible, patient-level splits are used, and the experiments report standard errors. The transfer-learning chapter is honest about the modest gains over an SVM and makes a reasonable robustness argument. As a dissertation, this is a fair compilation of the author's own prior work.\n\nThe soft spots are real and one is load-bearing. Section 1.2.3 claims a 92.5% AUC for detecting neurological injury, but the best result in the thesis's own Table 5.4 is 85.0±0.9 AUC. I could not find any model, feature set, domain-adaptation variant, patient split, or evaluation scheme that produces 92.5%. This is not a minor reporting inconsistency; it is the headline clinical claim. Second, the dataset counts do not add up: Section 5.3.1.2 says the curated birth subset has 1,108 audio samples (959 healthy, 149 neuro-injury), while Table 5.2 sums to 2,022 recordings (1,845 healthy, 177 neuro-injury). Either the text undercounts by roughly 900 samples or the table counts different units; the thesis doesn't say which. The label-noise concern around the modified Sarnat score is real but secondary to these traceability problems. The heavy reliance on the author's own preprints means the thesis itself does not present a new result beyond those earlier papers, but for a PhD dissertation that is normal; as an arXiv submission it reads as a compilation.\n\nThis paper is for people working on cry-based screening, audio model compression, and domain adaptation for medical audio. The CryCeleb dataset and tensorized RNN code are worth citing. But I would not rely on the 92.5% clinical claim until the authors reconcile it with Table 5.4 and clarify the data counts. A serious editor should not desk-reject this; it deserves peer review, but with clear instructions that the central clinical claim and the dataset description need to be fixed before acceptance.

Referee Report

4 major / 6 minor

Summary. This PhD thesis addresses medical audio analysis through four contributions: neural transfer learning from adult speech to infant-cry detection of perinatal asphyxia (Chapter 3); end-to-end tensor-train compression of recurrent networks (Chapter 4); self-supervised pre-training and domain adaptation on a new multi-center clinical cry database for neuro-injury and cry-trigger classification (Chapters 5-6); and the CryCeleb public dataset and challenge (Chapter 7). The central clinical claim in Section 1.2.3 is that the system detects neurological injury from newborn cries with an AUC of 92.5%. The thesis reports patient-level data splits, standard errors, open-source code for the tensorized RNNs, and a public cry dataset.

Significance. The individual contributions are meaningful if supported. The adult-speech transfer result in Chapter 3 is a useful baseline for low-resource cry analysis. The fully tensorized RNN in Chapter 4 offers a clean end-to-end parameterization with reported compression ratios of hundreds-fold and open-source code; the parameter-count derivation is sound. The SSL pipeline in Chapter 5 addresses a real data-scarcity problem, and the patient-level splits with repeated-seed standard errors are good experimental practice. The CryCeleb dataset and challenge are a genuine community resource. However, the headline clinical AUC is not traceable to the thesis's own tables, and the dataset description is internally inconsistent; these issues must be resolved before the clinical significance can be assessed.

major comments (4)

[1.2.3, 5.4.3] Section 1.2.3 states that the system detects neurological injury from newborns' cries 'with a high AUC of 92.5% (88.7% sensitivity at 80% specificity)', but the neuro-injury experiments in Chapter 5 peak at 85.0 +/- 0.9% AUC (Table 5.4, SSL + replay VGG, end-to-end). I could not find any section, table, equation, or figure that identifies the model, feature set, domain-adaptation variant, patient split, or per-hospital versus pooled evaluation that produces 92.5%. The headline number must be either traced to the reported experiments or removed and replaced with the reproducible result.
[5.3.1.2, Table 5.2] Section 5.3.1.2 says the curated labeled birth subset consists of 1,108 audio samples (959 healthy, 149 neuro-injury), but Table 5.2 reports train/validation/test totals of 1,845 healthy and 177 neuro-injury recordings, i.e., 2,022 recordings. Either the text undercounts by 914 samples or Table 5.2 uses a different unit (for example, segmented cry units). The thesis needs to state which unit is used, since the discrepancy makes it impossible to verify that the experiments were run on the described data.
[5.3.1.1, 5.3.1.2, 8.2] The clinical ground truth is a modified Sarnat score assigned by clinicians at birth and discharge, and Section 5.3.1.2 mentions that 'potentially mislabeled' recordings were removed. However, no inter-rater agreement, central adjudication, or validation against a reference standard such as MRI or standardized encephalopathy grading is reported. Because the central claim is a clinical screening claim, the thesis should provide label-quality evidence or explicitly frame all AUCs as benchmarked against chart-review Sarnat labels rather than against an independent diagnostic reference.
[6.5.5, Figure 6.10] The claim that target noise injection (TNI) requires neither labels nor cry recordings from the target domain needs qualification: Figure 6.10 varies the weight alpha and presumably selects it using some validation criterion. If that criterion uses target labels or a labeled target validation set, the 'unsupervised' characterization is overstated. The thesis should state the alpha-selection protocol and whether target-side labels are used at any point in the process.

minor comments (6)

[1.2.1, 3.2, 3.4.2, 5.2] The Chillanto database is described inconsistently: Sections 1.2.1 and 3.2 describe 1,389 recordings from 69 infants, Section 3.4.2 says 2,267 one-second cry samples from 88 infants, and Section 5.2 says the database contains only six patients diagnosed with birth asphyxia. These counts should be reconciled and tied to the exact subset used in Table 3.2.
[3.5.1] The text says a 16x12 MFCC representation is unrolled into a 168x1 feature vector, but 16 x 12 = 192; please correct the dimensions or describe the truncation that yields 168.
[1.3] The organization section lists Chapter 7 before Chapter 6, which conflicts with the actual chapter ordering in the thesis.
[4.2] The claim of 'state-of-the-art performance on the LibriSpeech dataset' should be qualified as state-of-the-art for the specific speaker-verification setup considered; as written it suggests a broader claim about LibriSpeech benchmarks.
[4.6, Table 4.1] Table 4.1 has malformed spacing in several time values (for example, '3 .70 +/- .19' and '2 .13 +/- .13'), which makes the table difficult to read and should be fixed.
[5.3.4, Figure 5.4] Section 5.3.4 says the model with the best validation score is repeated 10 times, while Figure 5.4 reports standard errors from five runs with different random seeds; the number of repeated runs should be stated consistently.

Circularity Check

1 steps flagged · score 4.0 of 10

Headline 92.5% neuro-injury AUC is carried by the author's own under-review preprint and is not derivable from the thesis's reported experiments (best Table 5.4 AUC is 85.0 ± 0.9).

self citation load bearing [Section 1.2.3 (Summary of research contributions), 'Addressing dataset bias and domain shift in medical audio'.]
"Our system extracts interpretable acoustic biomarkers that support clinical decisions and is able to accurately detect neurological injury from newborns' cries with a high AUC of 92.5% (88.7% sensitivity at 80% specificity). ... This work is presented in the papers: ... Onu, C. C., Latremouille, S. Gorin, A., Wang, J., Ekwochi, U., Ubuane, P. O., ... A cry for help: Early detection of brain injury in newborns. arXiv preprint arXiv:2310.08338, [164]."

The 92.5% AUC is presented as a central clinical contribution, but no section, table, or equation in the thesis derives it. The thesis's own neuro-injury classification experiments (Table 5.4) peak at 85.0 ± 0.9 AUC (SSL pre-training + replay, end-to-end fine-tuning). The only support offered for 92.5% is the citation of the author's own co-authored, under-review preprint [164], which is not reproduced, machine-checked, or independently validated in the thesis. The claim is therefore load-bearing on a self-citation rather than on the thesis's derivation chain.

full rationale

The methodological chapters are largely self-contained empirical work. Chapter 3 compares transfer-learning initializations against SVM and random-init baselines on the Chillanto test split; Chapter 4 derives tensor-train compression formulas and tests on MNIST and LibriSpeech; Chapter 5 uses held-out train/validation/test splits with patient-level separation and reports AUCs from repeated runs; Chapter 6 evaluates domain-adaptation methods on target test sets; Chapter 7 reports challenge baselines on a public dataset. None of these reduce by construction to their inputs, and none rely on a self-citation for the numerical result. The one load-bearing self-citation is the Introduction's 92.5% neuro-injury AUC, which is attributed to the author's own under-review preprint [164] and is not traceable to the experiments reported in the thesis (best in-thesis value: 85.0 ± 0.9, Table 5.4). This supports a moderate circularity score of 4, since the central headline claim rests on an unverified self-citation even though the thesis's individual chapters contain independent derivations. The separate inconsistency between the curated-subset count in Section 5.3.1.2 (1,108 samples) and Table 5.2 (2,022 recordings) is a verifiability/correctness issue, not a circularity issue, and is not counted toward the score.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claims rest on three domain assumptions rather than on invented physical entities. The free parameters are hyperparameters chosen by hand or validation, not derived quantities. The clinical claim depends on the reliability of the Sarnat score, which is not independently verified.

free parameters (3)

TT rank r = varies from 1 to 6 in experiments
In Chapter 4, the TT rank controls the trade-off between compression and accuracy; values are chosen per experiment, not derived.
Replay fraction for SSL cry adaptation = 50%
In Section 5.3.5, half of each batch is replayed from VGGSound; chosen heuristically.
Target noise weight alpha = not reported in excerpt
In TNI (Section 6.5.5), alpha controls the strength of target noise injection and is tuned on validation.

assumptions (3)

domain assumption Adult speech and infant cry share enough vocal-production structure that representations transfer.
Stated in Section 3.2 and tested empirically.
domain assumption Modified Sarnat score is a valid ground truth for neurological injury in newborns.
Used to label the Ubenwa dataset in Section 5.3.1.1; its reliability is not independently assessed.
ad hoc to paper Injecting target-domain noise into source representations simulates the domain shift without needing target labels.
Assumed by the TNI method in Section 6.5.5; no theoretical guarantee is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Making deep neural networks work for medical audio: representation, compression and domain adaptation." pith.science (2026). https://pith.science/paper/XLFXM6I6

@misc{pith2026250613970,
  author       = {Pith},
  title        = {Pith review of: Making deep neural networks work for medical audio: representation, compression and domain adaptation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XLFXM6I6}},
  note         = {Machine review of arXiv:2506.13970}
}

read the original abstract

This thesis addresses the technical challenges of applying machine learning to understand and interpret medical audio signals. The sounds of our lungs, heart, and voice convey vital information about our health. Yet, in contemporary medicine, these sounds are primarily analyzed through auditory interpretation by experts using devices like stethoscopes. Automated analysis offers the potential to standardize the processing of medical sounds, enable screening in low-resource settings where physicians are scarce, and detect subtle patterns that may elude human perception, thereby facilitating early diagnosis and treatment. Focusing on the analysis of infant cry sounds to predict medical conditions, this thesis contributes on four key fronts. First, in low-data settings, we demonstrate that large databases of adult speech can be harnessed through neural transfer learning to develop more accurate and robust models for infant cry analysis. Second, in cost-effective modeling, we introduce an end-to-end model compression approach for recurrent networks using tensor decomposition. Our method requires no post-hoc processing, achieves compression rates of several hundred-fold, and delivers accurate, portable models suitable for resource-constrained devices. Third, we propose novel domain adaptation techniques tailored for audio models and adapt existing methods from computer vision. These approaches address dataset bias and enhance generalization across domains while maintaining strong performance on the original data. Finally, to advance research in this domain, we release a unique, open-source dataset of infant cry sounds, developed in collaboration with clinicians worldwide. This work lays the foundation for recognizing the infant cry as a vital sign and highlights the transformative potential of AI-driven audio monitoring in shaping the future of accessible and affordable healthcare.

Figures

Figures reproduced from arXiv: 2506.13970 by the authors.

**Figure 3.1.** Conditions affecting infants and cry measures that each alters. [PITH_FULL_IMAGE:figures/full_fig_p060_3_1.png] view at source ↗

**Figure 3.2.** A physioacoustic model of the infant cry illustration how the central [PITH_FULL_IMAGE:figures/full_fig_p061_3_2.png] view at source ↗

**Figure 3.3.** Time-domain representation of a 5-sec cry signal. The y-axis represents [PITH_FULL_IMAGE:figures/full_fig_p065_3_3.png] view at source ↗

Figures from the paper (36 more)

**Figure 3.4.** Figure 3.4: (Top) The raw waveform of a 5-second cry signal. (Middle) The spectrum of of the first 25ms of the cry signal. The x-axis indicates the frequencies present and the y-axis shows the magnitudes of each. (Bottom) The spectrogram of the cry signal. The x-axis is time, th…

**Figure 3.5.** Figure 3.5: F0 or "Pitch" (blue line) and formants (red lines) of cry signal of a healthy baby in [PITH_FULL_IMAGE:figures/full_fig_p068_3_5.png]

**Figure 3.6.** Figure 3.6: Log mel-spectogram of the cry signal in Fig. [PITH_FULL_IMAGE:figures/full_fig_p069_3_6.png]

**Figure 3.7.** Figure 3.7: Mel-frequency cepstral coefficients (MFCC) of the cry signal in Fig. [PITH_FULL_IMAGE:figures/full_fig_p069_3_7.png]

**Figure 3.8.** Figure 3.8: Decision algorithm for voiced-unvoiced detection, proposed in [ [PITH_FULL_IMAGE:figures/full_fig_p070_3_8.png]

**Figure 3.9.** Figure 3.9: Neural transfer learning methodology. Weights from a pre-trained encoder [PITH_FULL_IMAGE:figures/full_fig_p077_3_9.png]

**Figure 3.10.** Figure 3.10: Performance of models under different noise conditions. [PITH_FULL_IMAGE:figures/full_fig_p080_3_10.png]

**Figure 3.11.** Figure 3.11: Audio length analysis highlighting the impact of using shorter amounts [PITH_FULL_IMAGE:figures/full_fig_p081_3_11.png]

**Figure 3.12.** Figure 3.12: Frequency response analysis of the relative importance of different Mel [PITH_FULL_IMAGE:figures/full_fig_p081_3_12.png]

**Figure 3.13.** Figure 3.13: Cumulative variance explained by all principal components (left) and the [PITH_FULL_IMAGE:figures/full_fig_p082_3_13.png]

**Figure 4.1.** Figure 4.1: Illustrated of tensorization process on GRU cell with hidden and input [PITH_FULL_IMAGE:figures/full_fig_p095_4_1.png]

**Figure 4.2.** Figure 4.2: Low dimensional UMAP visualization of embeddings from the TT-LSTM. [PITH_FULL_IMAGE:figures/full_fig_p101_4_2.png]

**Figure 4.3.** Figure 4.3: Illustration of regularization benefits of TT-RNNs. [PITH_FULL_IMAGE:figures/full_fig_p102_4_3.png]

**Figure 4.4.** Figure 4.4: Distribution of the norm of gradients of model parameters across 1000 [PITH_FULL_IMAGE:figures/full_fig_p103_4_4.png]

**Figure 5.1.** Figure 5.1: Protocol for the Ubenwa clinical study. Infant cry recordings are taken [PITH_FULL_IMAGE:figures/full_fig_p109_5_1.png]

**Figure 5.2.** Figure 5.2: Data collection interface. Patient information is added to the database [PITH_FULL_IMAGE:figures/full_fig_p110_5_2.png]

**Figure 5.3.** Figure 5.3: Summary of the proposed SSL-based training pipeline. [PITH_FULL_IMAGE:figures/full_fig_p112_5_3.png]

**Figure 5.4.** Figure 5.4: Performance using subsets of labeled neuro injury data in supervised fine [PITH_FULL_IMAGE:figures/full_fig_p122_5_4.png]

**Figure 5.5.** Figure 5.5: The 7 acoustic biomarkers of infant cry studied in this work. [PITH_FULL_IMAGE:figures/full_fig_p123_5_5.png]

**Figure 5.6.** Figure 5.6: Schematic representation of biomarker extraction. [PITH_FULL_IMAGE:figures/full_fig_p126_5_6.png]

**Figure 6.1.** Figure 6.1: In domain adaptation, we aim to build a cross-domain classifier that [PITH_FULL_IMAGE:figures/full_fig_p133_6_1.png]

**Figure 6.2.** Figure 6.2: Cross-hospital generalization results showing that model trained on only [PITH_FULL_IMAGE:figures/full_fig_p136_6_2.png]

**Figure 6.3.** Figure 6.3: Average distribution of spectral energy for healthy patients in hospital A [PITH_FULL_IMAGE:figures/full_fig_p137_6_3.png]

**Figure 6.4.** Figure 6.4: Average distribution of spectral energy for sick patients in hospital A (top) [PITH_FULL_IMAGE:figures/full_fig_p138_6_4.png]

**Figure 6.5.** Figure 6.5: Confusion matrix of a simple classifier trained to predict the originating [PITH_FULL_IMAGE:figures/full_fig_p139_6_5.png]

**Figure 6.6.** Figure 6.6: The pitch distribution of cry recordings from source and target domains [PITH_FULL_IMAGE:figures/full_fig_p142_6_6.png]

**Figure 6.7.** Figure 6.7: Performance of source → target adapted models on target test set [PITH_FULL_IMAGE:figures/full_fig_p157_6_7.png]

**Figure 6.8.** Figure 6.8: Impact of source → target adapted models on source test set [PITH_FULL_IMAGE:figures/full_fig_p157_6_8.png]

**Figure 6.9.** Figure 6.9: AUC in target domain as the amount of target domain noise is increased. [PITH_FULL_IMAGE:figures/full_fig_p158_6_9.png]

**Figure 6.10.** Figure 6.10: AUC in target domain for different values of [PITH_FULL_IMAGE:figures/full_fig_p159_6_10.png]

**Figure 7.1.** Figure 7.1: Histogram of cry sound durations. corresponding to each infant. However, cry sounds (expirations) collected within one recording session, tend to have similar acoustic characteristics. 7.4 CryCeleb 2023 Challenge CryCeleb 2023 was a two-month machine learning competi…

**Figure 7.2.** Figure 7.2: Number of infants per number of cry sounds. [PITH_FULL_IMAGE:figures/full_fig_p165_7_2.png]

**Figure 7.3.** Figure 7.3: CryCeleb challenge verification task. Given two recordings, pre [PITH_FULL_IMAGE:figures/full_fig_p165_7_3.png]

**Figure 7.4.** Figure 7.4: Verification scores for negative and positive pairs produced by the ECAPA [PITH_FULL_IMAGE:figures/full_fig_p169_7_4.png]

**Figure 7.5.** Figure 7.5: Verification scores for negative and positive pairs produced by the ECAPA [PITH_FULL_IMAGE:figures/full_fig_p170_7_5.png]

**Figure 7.6.** Figure 7.6: Equal error rates of CryCelb 2023 competition submissions. The public [PITH_FULL_IMAGE:figures/full_fig_p171_7_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

263 extracted references · 50 canonical work pages

[1]

A fully automated approach for baby cry signal segmentation and boundary detection of expi- ratory and inspiratory episodes

Lina Abou-Abbas, Chakib Tadj, and Hesam Alaie Fersaie. “A fully automated approach for baby cry signal segmentation and boundary detection of expi- ratory and inspiratory episodes”. In:The Journal of the Acoustical Society of America 142.3 (2017), pp. 1318–1331

2017
[2]

VCMNet: Weakly supervised learning for automatic infant vocalisation maturity analysis

Najla Al Futaisi et al. “VCMNet: Weakly supervised learning for automatic infant vocalisation maturity analysis”. In:International Conference on Multi- modal Interaction. 2019

2019
[3]

Tensor Decompositions for Learning Latent VariableModels

Animashree Anandkumar et al. “Tensor Decompositions for Learning Latent VariableModels”.In: Journal of Machine Learning Research15.1(2014),pp.2773– 2832

2014
[4]

Alzheimer’s disease: Communication patternsinfamilialandsporadicforms

A. Appell J. Kertesz and M. Fisman. “Alzheimer’s disease: Communication patternsinfamilialandsporadicforms”.In: Archives of Neurology39.10(1982), pp. 649–652

1982
[5]

On the differences between song and speech emotion recognition: Effect of feature sets, feature types, and classi- fiers

Bagus Tris Atmaja and Masato Akagi. “On the differences between song and speech emotion recognition: Effect of feature sets, feature types, and classi- fiers”. In: 2020 ieee region 10 conference (TENCON). IEEE. 2020, pp. 968– 972

2020
[6]

Do deep nets really need to be deep?

Jimmy Ba and Rich Caruana. “Do deep nets really need to be deep?” In: Advances in neural information processing systems. 2014, pp. 2654–2662. 150 BIBLIOGRAPHY 151

2014
[7]

A theory of learning from different domains

Shai Ben-David et al. “A theory of learning from different domains”. In:Ma- chine learning79 (2010), pp. 151–175

2010
[8]

Analysis of Representations for Domain Adaptation

Shai Ben-David et al. “Analysis of Representations for Domain Adaptation”. In: Advances in Neural Information Processing Systems (NIPS). Curran Asso- ciates, Inc., 2007, pp. 137–144

2007

Show all 263 references

[9]

Greedy Layer-Wise Training of Deep Networks

Yoshua Bengio et al. “Greedy Layer-Wise Training of Deep Networks”. In:Ad- vances in Neural Information Processing Systems. Vol. 19. Curran Associates, Inc. 2007, pp. 153–160

2007
[10]

Retrieved 3 August 2019

Paul Boersma and David Weenink.Praat: doing phonetics by computer [Com- puter program]. Retrieved 3 August 2019. Miscellaneous. 2019. url: http : //www.praat.org/

2019
[11]

A training algorithm for optimal margin classifiers

Bernhard E Boser, Isabelle M Guyon, and Vladimir N Vapnik. “A training algorithm for optimal margin classifiers”. In:Proceedings of the fifth annual workshop on Computational learning theory. ACM. 1992, pp. 144–152

1992
[12]

Application-independent evaluation of speaker detection

Niko Brümmer and Johan Du Preez. “Application-independent evaluation of speaker detection”. In:Computer Speech & Language20.2-3 (2006), pp. 230– 275

2006
[13]

CryCeleb: A Speaker Verification Dataset Based on Infant Cry Sounds

David Budaghyan et al. CryCeleb: A Speaker Verification Dataset Based on Infant Cry Sounds. 2023. arXiv:2305.00969 [cs.SD]. url: https://arxiv. org/abs/2305.00969

2023 arXiv
[14]

Dataset shift in machine learning

J Quinonero Candela et al. “Dataset shift in machine learning”. In:The MIT Press 1 (2009), p. 5

2009
[15]

Transfer Learning and Data Augmentation Tech- niques to the COVID-19 Identification Tasks in ComParE 2021

Edresson Casanova et al. “Transfer Learning and Data Augmentation Tech- niques to the COVID-19 Identification Tasks in ComParE 2021.” In:INTER- SPEECH. 2021. BIBLIOGRAPHY 152

2021
[16]

VGGSound: A Large-scale Audio-Visual Dataset

Honglie Chen et al. “VGGSound: A Large-scale Audio-Visual Dataset”. In: ICASSP. 2020

2020
[17]

VGGSound: A Large-scale Audio-Visual Dataset

Honglie Chen et al. VGGSound: A Large-scale Audio-Visual Dataset. 2020. arXiv: 2004.14368 [cs.CV]

2020 arXiv
[18]

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

Sanyuan Chen et al. “WavLM: Large-scale self-supervised pre-training for full stack speech processing”. In:IEEE Journal of Selected Topics in Signal Pro- cessing 16.6 (2022)

2022
[19]

A simple framework for contrastive learning of visual repre- sentations

Ting Chen et al. “A simple framework for contrastive learning of visual repre- sentations”. In:ICML. 2020

2020
[20]

Big self-supervised models are strong semi-supervised learn- ers

Ting Chen et al. “Big self-supervised models are strong semi-supervised learn- ers”. In:NeurIPS 33 (2020)

2020
[21]

Self-supervised learning based domain adaptation for robust speaker verification

Zhengyang Chen et al. “Self-supervised learning based domain adaptation for robust speaker verification”. In:ICASSP. 2021

2021
[22]

An exploration of parameter redundancy in deep networks with circulant projections

Yu Cheng et al. “An exploration of parameter redundancy in deep networks with circulant projections”. In:Proceedings of the IEEE International Confer- ence on Computer Vision. 2015, pp. 2857–2865

2015
[23]

Spectral analysis of infant cries and adultspeech

Anshu Chittora and Hemant A Patil. “Spectral analysis of infant cries and adultspeech”.In: International Journal of Speech Technology19(2016),pp.841– 856

2016
[24]

Learning Phrase Representations using RNN Encoder– Decoder for Statistical Machine Translation

Kyunghyun Cho et al. “Learning Phrase Representations using RNN Encoder– Decoder for Statistical Machine Translation”. In:Proceedings of the 2014 Con- ference on Empirical Methods in Natural Language Processing (EMNLP).2014, pp. 1724–1734

2014
[25]

VoxCeleb2: Deep Speaker Recog- nition

J. S. Chung, A. Nagrani, and A. Zisserman. “VoxCeleb2: Deep Speaker Recog- nition”. In:INTERSPEECH. 2018. BIBLIOGRAPHY 153

2018
[26]

Triplet Loss for Infant Cry Verification - CryCeleb2023 Solu- tion

Francesco Conti. Triplet Loss for Infant Cry Verification - CryCeleb2023 Solu- tion. https://github.com/conti748/cryceleb2023. Accessed: 2023-09-13

2023
[27]

An Algorithm for the Machine Cal- culation of Complex Fourier Series

James W. Cooley and John W. Tukey. “An Algorithm for the Machine Cal- culation of Complex Fourier Series”. In:Mathematics of Computation 19.90 (1965), pp. 297–301.doi: 10.1090/S0025-5718-1965-0178586-1

1965 doi
[28]

Support-Vector Networks

Corinna Cortes and Vladimir Vapnik. “Support-Vector Networks”. In: Ma- chine Learning 20.3 (1995), pp. 273–297. issn: 08856125. doi: 10.1023/A: 1022627411411. url: http://link.springer.com/10.1023/A:1022627411411

1995 doi
[29]

Cry analysis in in- fantsofnarcoticaddictedmothers

Michael J Corwin, Howard L Golub, and Margaret Potter. “Cry analysis in in- fantsofnarcoticaddictedmothers”.In: Pediatric Research21.4(1987),pp.180– 180

1987
[30]

The infant cry: what can it tell us?

Michael J Corwin, Barry M Lester, and Howard L Golub. “The infant cry: what can it tell us?” In:Current problems in Pediatrics26.9 (1996), pp. 313– 334

1996
[31]

Effects of in utero cocaine exposure on newborn acoustical cry characteristics

Michael J Corwin et al. “Effects of in utero cocaine exposure on newborn acoustical cry characteristics”. In:Pediatrics 89.6 (1992), pp. 1199–1203

1992
[32]

Approximation by superpositions of a sigmoidal function

George Cybenko. “Approximation by superpositions of a sigmoidal function”. In: Mathematics of Control, Signals, and Systems2.4 (1989), pp. 303–314

1989
[33]

Acoustical analysis of pain cries in neonates: Fundamental frequency

Raina P Daga and Anagha M Panditrao. “Acoustical analysis of pain cries in neonates: Fundamental frequency”. In:Int. J. Comput. Appl. Spec. Issue Electron. Inf. Commun. Eng ICEICE3 (2011), pp. 18–21

2011
[34]

Speech and swallowing prob- lems in multiple sclerosis

F. L. Darley, A. E. Aronson, and J. R. Brown. “Speech and swallowing prob- lems in multiple sclerosis”. In:Journal of Speech and Hearing Disorders37.4 (1972), pp. 485–495. BIBLIOGRAPHY 154

1972
[35]

Differential di- agnostic patterns of dysarthria

Frederic L. Darley, Arnold E. Aronson, and Joe R. Brown. “Differential di- agnostic patterns of dysarthria”. In:Journal of Speech and Hearing Research 12.2 (1969), pp. 246–269

1969
[36]

Comparison of Parametric Repre- sentations for Monosyllabic Word Recognition in Continuously Spoken Sen- tences

Steven B Davis and Paul Mermelstein. “Comparison of Parametric Repre- sentations for Monosyllabic Word Recognition in Continuously Spoken Sen- tences”. In:IEEE transactions on acoustics, speech, and signal processing28.4 (1980), pp. 357–366. url: https : / / pdfs . semanticsch...

1980
[37]

A Multilinear Singular Value Decomposition

Lieven De Lathauwer, Bart De Moor, and Joos Vandewalle. “A Multilinear Singular Value Decomposition”. In: SIAM Journal on Matrix Analysis and Applications 21.4 (2000), pp. 1253–1278

2000
[38]

Arcface: Additive angular margin loss for deep face recognition

Jiankang Deng et al. “Arcface: Additive angular margin loss for deep face recognition”. In:Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019, pp. 4690–4699

2019
[39]

Predicting parameters in deep learning

Misha Denil et al. “Predicting parameters in deep learning”. In:Advances in neural information processing systems. 2013, pp. 2148–2156

2013
[40]

ECAPA-TDNN: Emphasized channel attention, propagation and aggregation in TDNN based speaker verification

Brecht Desplanques et al. “ECAPA-TDNN: Emphasized channel attention, propagation and aggregation in TDNN based speaker verification”. In: IN- TERSPEECH (2020)

2020
[41]

Transfer learning of weakly labelled audio

Aleksandr Diment and Tuomas Virtanen. “Transfer learning of weakly labelled audio”. In:2017 ieee workshop on applications of signal processing to audio and acoustics (waspaa). IEEE. 2017, pp. 6–10

2017
[42]

Decaf: A deep convolutional activation feature for generic visual recognition

Jeff Donahue et al. “Decaf: A deep convolutional activation feature for generic visual recognition”. In:International conference on machine learning. PMLR. 2014, pp. 647–655. BIBLIOGRAPHY 155

2014
[43]

Emotional expression recognition using support vector ma- chines

Melanie Dumas. “Emotional expression recognition using support vector ma- chines”. In:Technical Report, Department of Computer Science and Engineer- ing. University of California, San Diego. 2001

2001
[44]

The Approximation of One Matrix by Another of Lower Rank

Carl Eckart and Gale Young. “The Approximation of One Matrix by Another of Lower Rank”. In:Psychometrika 1.3 (1936), pp. 211–218

1936
[45]

Finding Structure in Time

Jeffrey L. Elman. “Finding Structure in Time”. In: Cognitive Science 14.2 (1990), pp. 179–211

1990
[46]

Why does unsupervised pre-training help deep learn- ing?

Dumitru Erhan et al. “Why does unsupervised pre-training help deep learn- ing?” In:Journal of Machine Learning Research11.Feb (2010), pp. 625–660

2010
[47]

A Classification Model for Infant Cries with Hearing Impairment and Unilateral Cleft Lip and Palate

Tanja Etz, Henning Reetz, and Carla Wegener. “A Classification Model for Infant Cries with Hearing Impairment and Unilateral Cleft Lip and Palate”. In: Folia Phoniatrica et Logopaedica64.5 (2012), pp. 254–261.issn: 1421-9972. doi: 10.1159/000343994 . url: http://www.karger.com...

2012 doi
[48]

Opensmile: the munich versatile and fast open-source audio feature extractor

Florian Eyben et al. “Opensmile: the munich versatile and fast open-source audio feature extractor”. In:Proc. of the 18th ACM international conference on Multimedia. 2010

2010
[49]

TheGenevaminimalisticacousticparameterset(GeMAPS) for voice research and affective computing

FlorianEybenetal.“TheGenevaminimalisticacousticparameterset(GeMAPS) for voice research and affective computing”. In:IEEE transactions on affective computing 7.2 (2015), pp. 190–202

2015
[50]

Identification of infants’ cry motivation using spectro- grams

Gustavo Z Felipe et al. “Identification of infants’ cry motivation using spectro- grams”.In: International Conference on Systems, Signals and Image Processing (IWSSIP). IEEE. 2019

2019
[51]

Feynman, Robert B

Richard P. Feynman, Robert B. Leighton, and Matthew Sands.The Feynman Lectures on Physics, Vol. 1. Addison-Wesley, 1963. Chap. 49, pp. 1–11. BIBLIOGRAPHY 156

1963
[52]

Model-agnosticmeta-learning for fast adaptation of deep networks

ChelseaFinn,PieterAbbeel,andSergeyLevine.“Model-agnosticmeta-learning for fast adaptation of deep networks”. In:International conference on machine learning. PMLR. 2017, pp. 1126–1135

2017
[53]

Flanagan

James L. Flanagan. Speech Analysis, Synthesis and Perception. 2nd. Springer- Verlag, 1972

1972
[54]

Auditory Patterns

Harvey Fletcher. “Auditory Patterns”. In: Reviews of Modern Physics 12.1 (1940), pp. 47–65

1940
[55]

Loudness, its definition, measurement and calculation

Harvey Fletcher and W. A. Munson. “Loudness, its definition, measurement and calculation”. In:Journal of the Acoustical Society of America5.2 (1933), pp. 82–108

1933
[56]

Théorie Analytique de la Chaleur

Jean-Baptiste Joseph Fourier. Théorie Analytique de la Chaleur. Firmin Didot Père et Fils, 1822
[57]

Training batchnorm and only batchnorm: On the expressive power of random features in CNNs

Jonathan Frankle et al. “Training batchnorm and only batchnorm: On the expressive power of random features in CNNs”. In:ICLR (2020)

2020
[58]

Catastrophic forgetting in connectionist networks

Robert M. French. “Catastrophic forgetting in connectionist networks”. In: Trends in Cognitive Sciences3.4 (1999). issn: 1364-6613. doi: https://doi. org/10.1016/S1364-6613(99)01294-2. url: https://www.sciencedirect. com/science/article/pii/S1364661399012942

1999 doi
[59]

Neocognitron: A Self-organizing Neural Network Model for a Mechanism of Pattern Recognition Unaffected by Shift in Position

Kunihiko Fukushima. “Neocognitron: A Self-organizing Neural Network Model for a Mechanism of Pattern Recognition Unaffected by Shift in Position”. In: Biological Cybernetics36.4 (1980), pp. 193–202

1980
[60]

Support vector machines for speech recognition

Aarthi Ganapathiraju, Joseph Hamaker, and Joseph Picone. “Support vector machines for speech recognition”. In:Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)2 (1998), pp. 544–547. BIBLIOGRAPHY 157

1998
[61]

Unsupervised Domain Adaptation by Backpropagation

Yaroslav Ganin and Victor Lempitsky. “Unsupervised Domain Adaptation by Backpropagation”. In:International Conference on Machine Learning (ICML). 2015, pp. 1180–1189

2015
[62]

Domain-adversarial training of neural networks

Yaroslav Ganin et al. “Domain-adversarial training of neural networks”. In: The journal of machine learning research17.1 (2016), pp. 2096–2030

2016
[63]

Ultimate tensorization: compressing convolutional and fc layers alike

Timur Garipov et al. “Ultimate tensorization: compressing convolutional and fc layers alike”. In:NIPS 2016 workshop: Learning with Tensors: Why Now and How? (2016)

2016
[64]

Audio Set: An ontology and human-labeled dataset for audio events

Jort F. Gemmeke et al. “Audio Set: An ontology and human-labeled dataset for audio events”. In:Proc. IEEE ICASSP 2017. New Orleans, LA, 2017

2017
[65]

Understanding the difficulty of training deep feedforward neural networks

Xavier Glorot and Yoshua Bengio. “Understanding the difficulty of training deep feedforward neural networks”. In:Proceedings of the thirteenth interna- tional conference on artificial intelligence and statistics. 2010, pp. 249–256

2010
[66]

Deep sparse rectifier neu- ralnetworks

Xavier Glorot, Antoine Bordes, and Yoshua Bengio. “Deep sparse rectifier neu- ralnetworks”.In: Proceedings of the 14th International Conference on Artificial Intelligence and Statistics. 2011, pp. 315–323

2011
[67]

A Physioacoustic Model of the Infant Cry

Howard L. Golub and Michael J. Corwin. “A Physioacoustic Model of the Infant Cry”. In: Infant Crying. Boston, MA: Springer US, 1985, pp. 59–82. doi: 10.1007/978- 1- 4613- 2381- 5{\_}3 . url: http://link.springer. com/10.1007/978-1-4613-2381-5_3

1985 doi
[68]

A physioacoustic model of the infant cry

Howard L Golub and Michael J Corwin. “A physioacoustic model of the infant cry”. In: Infant crying: Theoretical and research perspectives. Springer, 1985, pp. 59–82

1985
[69]

SSAST: Self-supervised audio spectrogram transformer

Yuan Gong et al. “SSAST: Self-supervised audio spectrogram transformer”. In: Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 36. 10. 2022. BIBLIOGRAPHY 158

2022
[70]

Goodfellow et al.Generative Adversarial Networks

Ian J. Goodfellow et al.Generative Adversarial Networks. 2014. arXiv:1406. 2661 [stat.ML]

2014
[71]

Self-supervisedlearningforinfantcryanalysis

ArseniiGorinetal.“Self-supervisedlearningforinfantcryanalysis”.In: ICASSP Workshop SASB. IEEE. 2023

2023
[72]

Self-supervised learning for infant cry analysis

ArseniiGorin etal. Self-supervised learning for infant cry analysis. 2023.arXiv: 2305.01578 [cs.SD]

2023 arXiv
[73]

Semi-supervised Learning by Entropy Minimization

Yves Grandvalet and Yoshua Bengio. “Semi-supervised Learning by Entropy Minimization”. In:Advances in Neural Information Processing Systems. Ed. by L. Saul, Y. Weiss, and L. Bottou. Vol. 17. MIT Press, 2004.url: https:// proceedings.neurips.cc/paper_files/paper/2004/file/96f2b...

2004
[74]

A Kernel Two-Sample Test

Arthur Gretton et al. “A Kernel Two-Sample Test”. In:Journal of Machine Learning Research13 (2012), pp. 723–773

2012
[75]

Covariate Shift by Kernel Mean Matching

Arthur Gretton et al. “Covariate Shift by Kernel Mean Matching”. In:Dataset Shift in Machine Learning. The MIT Press, 2009, pp. 131–160

2009
[76]

Don’t stop pretraining: Adapt language models to domains and tasks

Suchin Gururangan et al. “Don’t stop pretraining: Adapt language models to domains and tasks”. In:ACL (2020)

2020
[77]

An assessment of par- alinguistic acoustic features for detection of Alzheimer’s dementia in sponta- neous speech

Fasih Haider, Sofia De La Fuente, and Saturnino Luz. “An assessment of par- alinguistic acoustic features for detection of Alzheimer’s dementia in sponta- neous speech”. In:IEEE Journal of Selected Topics in Signal Processing14.2 (2019), pp. 272–281

2019
[78]

Emotion recognition in low-resource settings: An evalua- tion of automatic feature selection methods

Fasih Haider et al. “Emotion recognition in low-resource settings: An evalua- tion of automatic feature selection methods”. In:Computer Speech & Language 65 (2021), p. 101119. BIBLIOGRAPHY 159

2021
[79]

Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding

Song Han, Huizi Mao, and William J Dally. “Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding”. In: arXiv preprint arXiv:1510.00149(2015)

2015 arXiv
[80]

On the Use of Windows for Harmonic Analysis with the Discrete Fourier Transform

Fredric J. Harris. “On the Use of Windows for Harmonic Analysis with the Discrete Fourier Transform”. In:Proceedings of the IEEE66.1 (1978), pp. 51– 83

1978
[81]

Mapping computer science research in Africa: using academic networking sites for assessing research activity

Matthew Harsh et al. “Mapping computer science research in Africa: using academic networking sites for assessing research activity”. In:Scientometrics 126 (2021), pp. 305–334

2021
[82]

Rethinkingimagenetpre-training

KaimingHe,RossGirshick,andPiotrDollár.“Rethinkingimagenetpre-training”. In: Proceedings of the IEEE/CVF international conference on computer vision. 2019, pp. 4918–4927

2019
[83]

Deep residual learning for image recognition

Kaiming He et al. “Deep residual learning for image recognition”. In:Proceed- ings of the IEEE conference on computer vision and pattern recognition. 2016, pp. 770–778

2016
[84]

Delving deep into rectifiers: Surpassing human-level perfor- mance on imagenet classification

Kaiming He et al. “Delving deep into rectifiers: Surpassing human-level perfor- mance on imagenet classification”. In:Proceedings of the IEEE international conference on computer vision. 2015, pp. 1026–1034

2015
[85]

Momentum Contrast for Unsupervised Visual Represen- tation Learning

Kaiming He et al. “Momentum Contrast for Unsupervised Visual Represen- tation Learning”. In:Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2020, pp. 9729–9738

2020
[86]

Streaming end-to-end speech recognition for mobile de- vices

Yanzhang He et al. “Streaming end-to-end speech recognition for mobile de- vices”. In: ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2019, pp. 6381–6385. BIBLIOGRAPHY 160

2019
[87]

End-to-endtext-dependentspeakerverification

GeorgHeigoldetal.“End-to-endtext-dependentspeakerverification”.In: 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2016, pp. 5115–5119

2016
[88]

Distilling the knowledge in a neural network

Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. “Distilling the knowledge in a neural network”. In:arXiv preprint arXiv:1503.02531(2015)

2015 arXiv
[89]

A Fast Learning Al- gorithm for Deep Belief Nets

Geoffrey E. Hinton, Simon Osindero, and Yee Whye Teh. “A Fast Learning Al- gorithm for Deep Belief Nets”. In:Neural Computation18.7 (2006), pp. 1527– 1554

2006
[90]

Long short-term memory

Sepp Hochreiter and Jürgen Schmidhuber. “Long short-term memory”. In: Neural computation9.8 (1997), pp. 1735–1780

1997
[91]

Long Short-Term Memory

Sepp Hochreiter and Jürgen Schmidhuber. “Long Short-Term Memory”. In: Neural Computation 9.8 (Nov. 1997), pp. 1735–1780. doi: 10 . 1162 / neco . 1997.9.8.1735 . url: http://www.mitpressjournals.org/doi/10.1162/ neco.1997.9.8.1735

1997
[92]

Neural networks and physical systems with emergent col- lective computational abilities

John J. Hopfield. “Neural networks and physical systems with emergent col- lective computational abilities”. In:Proceedings of the National Academy of Sciences 79.8 (1982), pp. 2554–2558

1982
[93]

Jeremy Howard and Sebastian Ruder.Universal Language Model Fine-tuning for Text Classification. 2018. arXiv:1801.06146 [cs.CL]

2018 arXiv
[94]

Correcting Sample Selection Bias by Unlabeled Data

Jiayuan Huang et al. “Correcting Sample Selection Bias by Unlabeled Data”. In: Advances in Neural Information Processing Systems (NeurIPS). Vol. 19. 2007, pp. 601–608

2007
[95]

Speech disorders in Parkinson’s disease: A clinical and phonetic study

M. Illman and K. W. Cross. “Speech disorders in Parkinson’s disease: A clinical and phonetic study”. In:Journal of Neurology, Neurosurgery, and Psychiatry 44.9 (1981), pp. 751–758. BIBLIOGRAPHY 161

1981
[96]

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

Sergey Ioffe and Christian Szegedy. “Batch normalization: Accelerating deep network training by reducing internal covariate shift”. In:International con- ference on machine learning. pmlr. 2015, pp. 448–456

2015
[97]

A review of infant cry analysis and classification

Chunyan Ji et al. “A review of infant cry analysis and classification”. In: EURASIP Journal on Audio, Speech, and Music Processing(2021)

2021
[98]

Principal component analysis

Ian Jolliffe. Principal component analysis. Springer, 2011

2011
[99]

Automated prediction of extubation success in extremely preterm infants: the APEX multicenter study

Lara J Kanbar et al. “Automated prediction of extubation success in extremely preterm infants: the APEX multicenter study”. In:Pediatric Research 93.4 (2023), pp. 1041–1049

2023
[100]

Undersampling and bagging of decision trees in the analysisofcardiorespiratorybehaviorforthepredictionofextubationreadiness in extremely preterm infants

Lara J Kanbar et al. “Undersampling and bagging of decision trees in the analysisofcardiorespiratorybehaviorforthepredictionofextubationreadiness in extremely preterm infants”. In:2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society...

2018
[101]

Large-scalevideoclassificationwithconvolutionalneu- ral networks

AndrejKarpathyetal. “Large-scalevideoclassificationwithconvolutionalneu- ral networks”. In:Proceedings of the IEEE conference on Computer Vision and Pattern Recognition. 2014, pp. 1725–1732

2014
[102]

A Comprehensive Review of Deep Learning: Ar- chitectures, Recent Advances and Applications

Waqas Khan and Imed Yairi. “A Comprehensive Review of Deep Learning: Ar- chitectures, Recent Advances and Applications”. In:Electronics 15.12 (2022), p. 755. doi: 10.3390/electronics11050755. url: https://www.mdpi.com/ 2078-2489/15/12/755

2022 doi
[103]

FrequentialCharacterizationofHealthy and Pathologic Newborns Cries

YasminaKheddacheandChakibTadj.“FrequentialCharacterizationofHealthy and Pathologic Newborns Cries”. In:American Journal of Biomedical Engi- neering 3.6 (2013), pp. 182–193. issn: 2163-1077. doi: 10 . 5923 / j . ajbe . 20130306.07. BIBLIOGRAPHY 162

2013
[104]

Tensorized Embedding Layers For Efficient Model Compression

Valentin Khrulkov et al. “Tensorized Embedding Layers For Efficient Model Compression”. In:arXiv preprint arXiv:1901.10787(2019)

2019 arXiv
[105]

Improving audio-language learning with mixgen and multi-leveltest-timeaugmentation

Eungbeom Kim et al. “Improving audio-language learning with mixgen and multi-leveltest-timeaugmentation”.In: arXiv preprint arXiv:2210.17143(2022)

2022 arXiv
[106]

Crepe: A convolutional representation for pitch esti- mation

Jong Wook Kim et al. “Crepe: A convolutional representation for pitch esti- mation”. In: 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2018, pp. 161–165

2018
[107]

Crepe: A Convolutional Representation for Pitch Es- timation

Jong Wook Kim et al. “Crepe: A Convolutional Representation for Pitch Es- timation”. In:2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2018, pp. 161–165.doi: 10.1109/ICASSP.2018. 8461329

2018 doi
[108]

Adam: A method for stochastic optimiza- tion

Diederik P Kingma and Jimmy Ba. “Adam: A method for stochastic optimiza- tion”. In:ICLR (2015)

2015
[109]

Audio for audio is better? An investigation on transfer learning models for heart sound classification

Tomoya Koike et al. “Audio for audio is better? An investigation on transfer learning models for heart sound classification”. In:2020 42nd Annual Inter- national Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). IEEE. 2020, pp. 74–77

2020
[110]

Tensor Decompositions and Applica- tions

Tamara G. Kolda and Brett W. Bader. “Tensor Decompositions and Applica- tions”. In:SIAM Review51.3 (2009), pp. 455–500

2009
[111]

PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition

Qiuqiang Kong et al. PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition. 2020. arXiv:1912.10211 [cs.SD]

2020 arXiv
[112]

PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition

Qiuqiang Kong et al. “PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition”. In: IEEE/ACM Trans. Audio, Speech and Lang. Proc.28 (2020). issn: 2329-9290. doi: 10.1109/TASLP.2020.3030497. url: https://doi.org/10.1109/TASLP.2020.3030497. BIBLIOGRAPHY 163

2020
[113]

Assessment of infant cry: acoustic cry analysis and parental perception

Linda L LaGasse, A Rebecca Neal, and Barry M Lester. “Assessment of infant cry: acoustic cry analysis and parental perception”. In:Mental retardation and developmental disabilities research reviews11.1 (2005), pp. 83–93

2005
[114]

Assessmentofinfant cry:Acousticcryanalysisandparentalperception

LindaL.LaGasse,A.RebeccaNeal,andBarryM.Lester.“Assessmentofinfant cry:Acousticcryanalysisandparentalperception”.In: Mental Retardation and Developmental Disabilities Research Reviews11.1 (Feb. 2005), pp. 83–93.issn: 1080-4013. doi: 10.1002/mrdd.20050 . url: http://doi.wiley.co...

2005 doi
[115]

Characteristics of newborns with hypoxic is- chemic encephalopathy treated in NICUs at three different income-level coun- tries

Samantha Latremouille et al. “Characteristics of newborns with hypoxic is- chemic encephalopathy treated in NICUs at three different income-level coun- tries”. In:Congress of Joint European Neonatal Societies(2023)

2023
[116]

4 million neonatal deaths: when? Where? Why?

Joy E Lawn, Simon Cousens, and Jelka Zupan. “4 million neonatal deaths: when? Where? Why?” In:The lancet365.9462 (2005), pp. 891–900

2005
[117]

TheMNISTdatabase of handwritten digits, 1998

YannLeCun,CorinnaCortes,andChristopherJCBurges.“TheMNISTdatabase of handwritten digits, 1998”. In: URL http://yann. lecun. com/exdb/mnist 10.34 (1998), p. 14

1998
[118]

Backpropagation Applied to Handwritten Zip Code Recog- nition

Yann LeCun et al. “Backpropagation Applied to Handwritten Zip Code Recog- nition”. In:Neural Computation1.4 (1989), pp. 541–551

1989
[119]

Gradient-based learning applied to document recognition

Yann LeCun et al. “Gradient-based learning applied to document recognition”. In: Proceedings of the IEEE86.11 (1998), pp. 2278–2324

1998
[120]

Effects of marijuana use during preg- nancy on newborn cry

Barry M Lester and Melanie Dreher. “Effects of marijuana use during preg- nancy on newborn cry”. In:Child development (1989), pp. 765–771

1989
[121]

Colic for developmentalists

Barry M Lester et al. “Colic for developmentalists”. In:Infant Mental Health Journal 11.4 (1990), pp. 321–333. BIBLIOGRAPHY 164

1990
[122]

The maternal lifestyle study: effects of substance ex- posure during pregnancy on neurodevelopmental outcome in 1-month-old in- fants

Barry M Lester et al. “The maternal lifestyle study: effects of substance ex- posure during pregnancy on neurodevelopmental outcome in 1-month-old in- fants”. In:Pediatrics 110.6 (2002), pp. 1182–1192

2002
[123]

Deepspeaker:anend-to-endneuralspeakerembeddingsystem

ChaoLietal.“Deepspeaker:anend-to-endneuralspeakerembeddingsystem”. In: arXiv preprint arXiv:1705.02304(2017)

2017 arXiv
[124]

Pruningfiltersforefficientconvnets

HaoLietal.“Pruningfiltersforefficientconvnets”.In: arXiv preprint arXiv:1608.08710 (2016)

2016 arXiv
[125]

A waveform-feature dual branch acoustic embedding net- work for emotion recognition

Jeng-Lin Li et al. “A waveform-feature dual branch acoustic embedding net- work for emotion recognition”. In:Frontiers in Computer Science2 (2020), p. 13

2020
[126]

A review of applications in federated learning

Li Li et al. “A review of applications in federated learning”. In:Computers & Industrial Engineering149 (2020), p. 106854

2020
[127]

Revisiting batch normalization for practical domain adap- tation

Yanghao Li et al. “Revisiting batch normalization for practical domain adap- tation”. In:arXiv preprint arXiv:1603.04779(2016)

2016 arXiv
[128]

Spectographic analysis of vocal response to pain stimuli in infants with Down’s syndrome

J Lind et al. “Spectographic analysis of vocal response to pain stimuli in infants with Down’s syndrome”. In:Developmental Medicine & Child Neurology12.4 (1970), pp. 478–486

1970
[129]

Infant cry language analysis and recognition: an experi- mental approach

Lichuan Liu et al. “Infant cry language analysis and recognition: an experi- mental approach”. In:IEEE/CAA Journal of Automatica Sinica6.3 (2019)

2019
[130]

Unsupervised Domain Adaptation with Residual Trans- fer Networks

Mingsheng Long et al. Unsupervised Domain Adaptation with Residual Trans- fer Networks. 2017. arXiv:1602.04433 [cs.LG]

2017 arXiv
[131]

A unified approach to interpreting model predictions

Scott Lundberg. “A unified approach to interpreting model predictions”. In: arXiv preprint arXiv:1705.07874(2017). BIBLIOGRAPHY 165

2017 arXiv
[132]

Multi-corpus experiment on continuous speech emotion recognition: convolution or recurrence?

Manon Macary et al. “Multi-corpus experiment on continuous speech emotion recognition: convolution or recurrence?” In:Speech and Computer: 22nd In- ternational Conference, SPECOM 2020, St. Petersburg, Russia, October 7–9, 2020, Proceedings 22. Springer. 2020, pp. 304–314

2020
[133]

Principal components analysis (PCA)

Andrzej Maćkiewicz and Waldemar Ratajczak. “Principal components analysis (PCA)”. In:Computers & Geosciences19.3 (1993), pp. 303–342

1993
[134]

A Robust Tool for Newborn Infant Cry Analysis

Claudia Manfredi, Valentina Tocchioni, and Leonardo Bocchi. “A Robust Tool for Newborn Infant Cry Analysis”. In:2006 International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE, Aug. 2006, pp. 509–

2006
[135]

A Survey on Self-supervised Pre-training for Sequential Transfer Learning in Neural Networks

Huanru Henry Mao. “A Survey on Self-supervised Pre-training for Sequential Transfer Learning in Neural Networks”. In:arXiv preprint arXiv:2007.00800 (2020)

2020 arXiv
[136]

The SIFT algorithm for fundamental frequency estimation

J. Markel. “The SIFT algorithm for fundamental frequency estimation”. In: IEEE Transactions on Audio and Electroacoustics20.5 (Dec. 1972), pp. 367–

1972
[137]

Pearson/Prentice Hall Upper Saddle River, 2009

James H Martin and Daniel Jurafsky.Speech and language processing: An in- troduction to natural language processing, computational linguistics, and speech recognition. Pearson/Prentice Hall Upper Saddle River, 2009

2009
[138]

Supervised and Unsupervised Learning of Audio Representations for Music Understanding

Matthew C McCallum et al. “Supervised and Unsupervised Learning of Audio Representations for Music Understanding”. In:ISMIR (2022)

2022
[139]

A Logical Calculus of the Ideas Im- manent in Nervous Activity

Warren S. McCulloch and Walter Pitts. “A Logical Calculus of the Ideas Im- manent in Nervous Activity”. In:The Bulletin of Mathematical Biophysics5.4 (1943), pp. 115–133. BIBLIOGRAPHY 166

1943
[140]

librosa: Audio and music signal analysis in python

Brian McFee et al. “librosa: Audio and music signal analysis in python.” In: SciPy. 2015, pp. 18–24

2015
[141]

UMAP: Uniform mani- fold approximation and projection for dimension reduction

Leland McInnes, John Healy, and James Melville. “UMAP: Uniform mani- fold approximation and projection for dimension reduction”. In:arXiv preprint arXiv:1802.03426 (2018)

2018 arXiv
[142]

The Speakers in the Wild (SITW) Speaker Recogni- tion Database

Mitchell McLaren et al. “The Speakers in the Wild (SITW) Speaker Recogni- tion Database.” In:Interspeech. 2016, pp. 818–822

2016
[143]

Functions of positive and negative type and their connection with the theory of integral equations

James Mercer. “Functions of positive and negative type and their connection with the theory of integral equations”. In:Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences209 (1909), pp. 415–446

1909
[144]

Sound spectrographic cry anal- ysis of infant with bacterial meningitis

K Michelsson, P Sirvio, and O Wasz-Hockert. “Sound spectrographic cry anal- ysis of infant with bacterial meningitis”. In:Dev Med Child Neurol19 (1977), pp. 309–315

1977
[146]

CryAnalysesOfSymptomlessLowBirthWeightNeonates And Of Asphyxiated Newborn Infants

KatarinaMichelsson.“CryAnalysesOfSymptomlessLowBirthWeightNeonates And Of Asphyxiated Newborn Infants”. In:Acta Paediatrica60.S216 (1971), pp. 9 –45. doi: 10 . 1111 / j . 1651 - 2227 . 1971 . tb05679 . x. url: http : / / onlinelibrary.wiley.com/doi/10.1111/j.1651-2227.1971.tb0...

1971
[147]

Phonation in the newborn, infant cry

Katarina Michelsson and Oliver Michelsson. “Phonation in the newborn, infant cry”. In:International Journal of Pediatric Otorhinolaryngology49.SUPPL. 1 (1999), pp. 297–301.issn: 01655876. doi: 10.1016/S0165-5876(99)00180-9. BIBLIOGRAPHY 167

1999 doi
[148]

Pain cry in full- term asphyxiated newborn infants correlated with late findings

Katarina Michelsson, Pertti Sirviö, and Ole Wasz-Höckert. “Pain cry in full- term asphyxiated newborn infants correlated with late findings”. In:Acta Pæ- diatrica 66.5 (1977), pp. 611–616

1977
[149]

Sound spectro- graphic cry analysis of infants with bacterial meningitis

Katarina Michelsson, Pertti Sirviö. A, and Ole Wasz-Höckert. “Sound spectro- graphic cry analysis of infants with bacterial meningitis”. In:Developmental Medicine & Child Neurology19.3 (1977), pp. 309–315

1977
[150]

Cry characteristics of 172 healthy 1-to 7-day-old infants

Katarina Michelsson et al. “Cry characteristics of 172 healthy 1-to 7-day-old infants”. In:Folia phoniatrica et logopaedica54.4 (2002), pp. 190–200

2002
[151]

Sound spectrographic cry analysis in neonatal di- agnostics. An evaluative study

Katarina Michelsson et al. “Sound spectrographic cry analysis in neonatal di- agnostics. An evaluative study”. In:Journal of phonetics10.1 (1982), pp. 79– 88

1982
[152]

Measurement of the fundamental period of speech using a delay line

RL Miller and ES Weibel. “Measurement of the fundamental period of speech using a delay line”. In:The Journal of the Acoustical Society of America28.4 (1956), pp. 761–761

1956
[153]

Self-supervised speech representation learning: A review

Abdelrahman Mohamed et al. “Self-supervised speech representation learning: A review”. In:IEEE Journal of Selected Topics in Signal Processing(2022)

2022
[154]

Where do poor women in developing countries give birth? A multi-country analysis of demographic and health survey data

Dominic Montagu et al. “Where do poor women in developing countries give birth? A multi-country analysis of demographic and health survey data”. In: PloS one6.2 (2011), e17155

2011
[155]

Brian C. J. Moore. An Introduction to the Psychology of Hearing. 6th. Brill, 2012

2012
[156]

A scoping reviewoftransferlearningresearchonmedicalimageanalysisusingImageNet

Mohammad Amin Morid, Alireza Borjali, and Guilherme Del Fiol. “A scoping reviewoftransferlearningresearchonmedicalimageanalysisusingImageNet”. In: Computers in biology and medicine128 (2021), p. 104115. BIBLIOGRAPHY 168

2021
[157]

Data augmentation approaches for improving animal audio classification

Loris Nanni, Gianluca Maguolo, and Michelangelo Paci. “Data augmentation approaches for improving animal audio classification”. In:Ecological Informat- ics 57 (2020), p. 101084.issn: 1574-9541. doi: https://doi.org/10.1016/ j.ecoinf.2020.101084. url: https://www.sciencedirect....

2020
[158]

Short-time spectrum and “cepstrum

A Michael Noll. “Short-time spectrum and “cepstrum” techniques for vocal- pitch detection”. In: The Journal of the Acoustical Society of America36.2 (1964), pp. 296–302

1964
[159]

Tensorizingneuralnetworks

AlexanderNovikovetal.“Tensorizingneuralnetworks”.In: Advances in Neural Information Processing Systems. 2015, pp. 442–450

2015
[160]

Theeffectsofmaternalalcoholconsumptionandcigarette smoking during pregnancy on acoustic cry analysis

JKevinNugentetal.“Theeffectsofmaternalalcoholconsumptionandcigarette smoking during pregnancy on acoustic cry analysis”. In:Child Development 67.4 (1996), pp. 1806–1815

1996
[161]

Ubenwa: Cry-based Diagnosis of Birth Asphyxia

CC Onu et al. “Ubenwa: Cry-based Diagnosis of Birth Asphyxia”. In:NIPS Workshop on Machine Learning for the Developing World(2017)

2017
[162]

Harnessing infant cry for swift, cost-effective diagnosis of Peri- natal Asphyxia in low-resource settings

Charles C Onu. “Harnessing infant cry for swift, cost-effective diagnosis of Peri- natal Asphyxia in low-resource settings”. In:2014 IEEE Canada International Humanitarian Technology Conference-(IHTC). IEEE. 2014, pp. 1–4

2014
[163]

AFullyTensorizedRecur- rent Neural Network

CharlesC.Onu,JacobE.Miller,andDoinaPrecup.“AFullyTensorizedRecur- rent Neural Network”. In:CoRR abs/2010.04196 (2020). arXiv:2010.04196. url: https://arxiv.org/abs/2010.04196

2020 arXiv
[164]

Onu et al

Charles C. Onu et al. A cry for help: Early detection of brain injury in new- borns. 2023. arXiv:2310.08338 [eess.AS]

2023 arXiv
[165]

Onu et al

Charles C. Onu et al. Learning domain-invariant classifiers for infant cry sounds. 2023. arXiv: 2312 . 00231 [eess.AS]. url: https : / / arxiv . org / abs/2312.00231. BIBLIOGRAPHY 169

2023 arXiv
[166]

Neural transfer learning for cry-based diagnosis of peri- natal asphyxia

Charles C Onu et al. “Neural transfer learning for cry-based diagnosis of peri- natal asphyxia”. In:INTERSPEECH (2019)

2019
[167]

Onu et al

Charles C. Onu et al. Neural Transfer Learning for Cry-based Diagnosis of Perinatal Asphyxia. 2020. arXiv:1906.10199 [cs.LG]

2020 arXiv
[168]

Ubenwa: Cry-based diagnosis of birth asphyxia

Charles C Onu et al. “Ubenwa: Cry-based diagnosis of birth asphyxia”. In: arXiv preprint arXiv:1711.06405(2017)

2017 arXiv
[169]

Oppenheim and Ronald W

Alan V. Oppenheim and Ronald W. Schafer.Discrete-Time Signal Processing. 2nd. Prentice Hall, 1999

1999
[170]

Learning and transferring mid-level image representa- tions using convolutional neural networks

Maxime Oquab et al. “Learning and transferring mid-level image representa- tions using convolutional neural networks”. In:Proceedings of the IEEE con- ference on computer vision and pattern recognition. 2014, pp. 1717–1724

2014
[171]

Application of Pattern Recognition Techniques to the Classification of Full-Term and Preterm Infant Cry

Silvia Orlandi et al. “Application of Pattern Recognition Techniques to the Classification of Full-Term and Preterm Infant Cry”. In:Journal of Voice30.6 (Nov. 2016), pp. 656–663. issn: 0892-1997. doi: 10.1016/J.JVOICE.2015. 08.007. url: https://www.sciencedirect.com/science/ar...

2016 doi
[172]

Tensor-Train Decomposition

Ivan V Oseledets. “Tensor-Train Decomposition”. In:SIAM Journal on Scien- tific Computing33.5 (2011), pp. 2295–2317

2011
[173]

Infant cry classification by using different deep neural net- work models and hand-crafted features

Turgut Ozseven. “Infant cry classification by using different deep neural net- work models and hand-crafted features”. In:Biomedical Signal Processing and Control 83 (2023)

2023
[174]

A Survey on Transfer Learning

Sinno Jialin Pan and Qiang Yang. “A Survey on Transfer Learning”. In:IEEE Transactions on Knowledge and Data Engineering22.10 (Oct. 2010), pp. 1345–

2010
[175]

Librispeech: an asr corpus based on public domain audio books

Vassil Panayotov et al. “Librispeech: an asr corpus based on public domain audio books”. In: 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2015, pp. 5206–5210

2015
[176]

Defining and distinguishing infant behavioral states using acoustic cry analysis: is colic painful?

Joanna J Parga et al. “Defining and distinguishing infant behavioral states using acoustic cry analysis: is colic painful?” In:Pediatric research87.3 (2020)

2020
[177]

SpecAugment: A simple data augmentation method for automatic speech recognition

Daniel S Park et al. “SpecAugment: A simple data augmentation method for automatic speech recognition”. In:INTERSPEECH (2019)

2019
[178]

Auditory identification of pain cry signals of young infants in pathological conditions and its sound spectrographic basis

TJ Partanen et al. “Auditory identification of pain cry signals of young infants in pathological conditions and its sound spectrographic basis”. In: Annales Paediatriae Fenniae. Vol. 13. 2. 1967, pp. 56–63

1967
[179]

PyTorch: An Imperative Style, High-Performance Deep Learning Library

Adam Paszke et al. “PyTorch: An Imperative Style, High-Performance Deep Learning Library”. In:Advances in Neural Information Processing Systems 32. Ed. by H. Wallach et al. Curran Associates, Inc., 2019, pp. 8024–8035.url: http://papers.neurips.cc/paper/9015-pytorch-an-imperat...

2019
[180]

Unsupervised domain adaptation for medical imag- ing segmentation with self-ensembling

Christian S Perone et al. “Unsupervised domain adaptation for medical imag- ing segmentation with self-ensembling”. In:NeuroImage 194 (2019), pp. 1–11

2019
[181]

Classification of infant cry vocalizations using artificial neural networks (ANNs)

Marco Petroni et al. “Classification of infant cry vocalizations using artificial neural networks (ANNs)”. In:International Conference on Acoustics, Speech, and Signal Processing. Vol. 5. IEEE, 1995, pp. 3475–3478.isbn: 0-7803-2431-5. doi: 10.1109/ICASSP.1995.479734 . url: htt...

1995
[182]

ESC: Dataset for Environmental Sound Classification

Karol J. Piczak. “ESC: Dataset for Environmental Sound Classification”. In: Proceedings of the 23rd Annual ACM Conference on Multimedia. Brisbane, Australia: ACM Press, Oct. 13, 2015, pp. 1015–1018.isbn: 978-1-4503-3459-4. doi: 10.1145/2733373.2806390. url: http://dl.acm.org/c...

2015
[183]

Analyzing infant cries using a committee of neural networks in order to detect hypoxia related disorder

Mannes Poel and Taco Ekkel. “Analyzing infant cries using a committee of neural networks in order to detect hypoxia related disorder”. In:International journal on artificial intelligence tools15.03 (2006), pp. 397–410

2006
[184]

Can we trust deep learning based diagnosis? the impact of domain shift in chest radiograph classification

Eduardo HP Pooch, Pedro Ballester, and Rodrigo C Barros. “Can we trust deep learning based diagnosis? the impact of domain shift in chest radiograph classification”. In: Thoracic Image Analysis: Second International Workshop, TIA 2020, Held in Conjunction with MICCAI 2020, Lim...

2020
[185]

Prospective research on infants with mild encephalopa- thy: the PRIME study

C Prempunpong et al. “Prospective research on infants with mild encephalopa- thy: the PRIME study”. In:Journal of Perinatology38.1 (2018), pp. 80–85

2018
[186]

Induction of decision trees

J. Ross Quinlan. “Induction of decision trees”. In:Machine learning1 (1986), pp. 81–106

1986
[187]

https://github.com/viksit-siddhant/ CryCeleb23

Siddhant Rai Viksit and Vinayak Abrol.Cross-Caps Lab’s Wining System Sub- mission for CryCeleb23 Challenge. https://github.com/viksit-siddhant/ CryCeleb23. Accessed: 2023-09-13

2023
[188]

TensorMap: Lidar-Based Topological Mapping and Localization via Tensor Decomposi- tions

Sirisha Rambhatla, Nikos D. Sidiropoulos, and Jarvis Haupt. “TensorMap: Lidar-Based Topological Mapping and Localization via Tensor Decomposi- tions”. In:arXiv preprint arXiv:1902.10226(2019)

2019 arXiv
[189]

Xnor-net: Imagenet classification using binary convolutional neural networks

Mohammad Rastegari et al. “Xnor-net: Imagenet classification using binary convolutional neural networks”. In:European conference on computer vision. Springer. 2016, pp. 525–542

2016
[191]

A System for the Processing of Infant Cry to Recognize Pathologies in Recently Born Babies with Neural Networks

Orion F Reyes-Galaviz and Carlos Alberto Reyes-Garcia. “A System for the Processing of Infant Cry to Recognize Pathologies in Recently Born Babies with Neural Networks”. In: SPECOM’2004: 9th Conference on Speech and BIBLIOGRAPHY 172 Computer. St Petersburg, Russia, 2004.url: h...

2004
[192]

A system for the processing of infant cry to recognize pathologies in recently born babies with neural networks

Orion F Reyes-Galaviz and Carlos Alberto Reyes-Garcia. “A system for the processing of infant cry to recognize pathologies in recently born babies with neural networks”. In:SPECOM. 2004

2004
[193]

Evolutionary-Neural System to Classify Infant Cry Units for Pathologies Identification in Recently Born Babies

Orion Fausto Reyes-Galaviz, Sergio Daniel Cano-Ortiz, and Carlos Alberto Reyes-García. “Evolutionary-Neural System to Classify Infant Cry Units for Pathologies Identification in Recently Born Babies”. In:2008 Seventh Mexican International Conference on Artificial Intelligence....

2008
[194]

" Why should i trust you?

Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. “" Why should i trust you?" Explaining the predictions of any classifier”. In:Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining. 2016, pp. 1135–1144

2016
[195]

Catastrophic forgetting, rehearsal and pseudorehearsal

Anthony Robins. “Catastrophic forgetting, rehearsal and pseudorehearsal”. In: Connection Science7.2 (1995), pp. 123–146

1995
[196]

The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain

Frank Rosenblatt. “The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain”. In:Psychological Review65.6 (1958), pp. 386–408

1958
[197]

Rossing, F

Thomas D. Rossing, F. Richard Moore, and Paul A. Wheeler.The Science of Sound. 3rd. Addison-Wesley, 2002

2002
[198]

Learning representations by back-propagating errors

David E. Rumelhart, Geoffrey E. Hinton, and Ronald J. Williams. “Learning representations by back-propagating errors”. In:Nature 323 (1986), pp. 533– 536. BIBLIOGRAPHY 173

1986
[199]

Optimization of principal component analysis and sup- port vector machine for the recognition of infant cry with asphyxia

Rohilah Sahak et al. “Optimization of principal component analysis and sup- port vector machine for the recognition of infant cry with asphyxia”. In:In- ternational Journal of Computers and Applications35.3 (2013), pp. 99–107

2013
[200]

Low-rank matrix factorization for deep neural network training with high-dimensional output targets

Tara N Sainath et al. “Low-rank matrix factorization for deep neural network training with high-dimensional output targets”. In:2013 IEEE international conference on acoustics, speech and signal processing. IEEE. 2013, pp. 6655– 6659

2013
[201]

Neonatal encephalopathy following fetal distress: a clinical and electroencephalographic study

Harvey B Sarnat and Margaret S Sarnat. “Neonatal encephalopathy following fetal distress: a clinical and electroencephalographic study”. In:Archives of neurology 33.10 (1976)

1976
[202]

Paralinguisticsinspeechandlanguage—State- of-the-art and the challenge

BjörnSchullerandAntonBatliner.“Paralinguisticsinspeechandlanguage—State- of-the-art and the challenge”. In:Computer Speech & Language27.1 (2013), pp. 4–39

2013
[203]

The interspeech 2016 computational paralinguistics chal- lenge: Deception, sincerity & native language

Björn Schuller et al. “The interspeech 2016 computational paralinguistics chal- lenge: Deception, sincerity & native language”. In:INTERSPEECH. 2016

2016
[204]

CNN features off-the-shelf: an astounding baseline for recognition

Ali Sharif Razavian et al. “CNN features off-the-shelf: an astounding baseline for recognition”. In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops. 2014, pp. 806–813

2014
[205]

Wasserstein distance guided representation learning for do- main adaptation

Jian Shen et al. “Wasserstein distance guided representation learning for do- main adaptation”. In:Proceedings of the AAAI Conference on Artificial Intel- ligence. Vol. 32. 1. 2018

2018
[206]

Deep unsupervised domain adaptation with time series sensor data: A survey

Yongjie Shi, Xianghua Ying, and Jinfa Yang. “Deep unsupervised domain adaptation with time series sensor data: A survey”. In:Sensors 22.15 (2022), p. 5507. BIBLIOGRAPHY 174

2022
[207]

Improving predictive inference under covariate shift by weighting the log-likelihood function

Hidetoshi Shimodaira. “Improving predictive inference under covariate shift by weighting the log-likelihood function”. In:Journal of Statistical Planning and Inference 90.2 (2000), pp. 227–244

2000
[208]

Deep convolutional neural networks for computer-aided detection:CNNarchitectures,datasetcharacteristicsandtransferlearning

Hoo-Chang Shin et al. “Deep convolutional neural networks for computer-aided detection:CNNarchitectures,datasetcharacteristicsandtransferlearning”.In: IEEE transactions on medical imaging35.5 (2016), pp. 1285–1298

2016
[209]

Very deep convolutional networks for large-scale image recognition

Karen Simonyan. “Very deep convolutional networks for large-scale image recognition”. In:arXiv preprint arXiv:1409.1556(2014)

2014 arXiv
[210]

https : / / huggingface

SpeechBrain VoxCeleb model. https : / / huggingface . co / speechbrain / spkrec-ecapa-voxceleb. Accessed: 2023-04-30

2023
[211]

Contrastive learning of musical representations

Janne Spijkervet and John Ashley Burgoyne. “Contrastive learning of musical representations”. In:ISMIR (2021)

2021
[212]

A closer look at domain shift for deep learning in histopathol- ogy. arXiv

K Stacke et al. “A closer look at domain shift for deep learning in histopathol- ogy. arXiv”. In:arXiv preprint arXiv:1909.1157510 (2019)

2019 arXiv
[213]

A Scale for the Measurement of the Psychological Magnitude Pitch

S.S. Stevens, J. Volkmann, and E.B. Newman. “A Scale for the Measurement of the Psychological Magnitude Pitch”. In:Journal of the Acoustical Society of America8.3 (1937), pp. 185–190

1937
[214]

Return of frustratingly easy domain adaptation

Baochen Sun, Jiashi Feng, and Kate Saenko. “Return of frustratingly easy domain adaptation”. In:Proceedings of the AAAI conference on artificial in- telligence. Vol. 30. 1. 2016

2016
[215]

Unsupervised Domain Adapta- tion through Self-Supervision

Baochen Sun, Jiashi Feng, and Kate Saenko. “Unsupervised Domain Adapta- tion through Self-Supervision”. In:Advances in Neural Information Processing Systems (NeurIPS). Vol. 32. 2019, pp. 6827–6837

2019
[216]

Deep coral: Correlation alignment for deep domain adaptation

Baochen Sun and Kate Saenko. “Deep coral: Correlation alignment for deep domain adaptation”. In: Computer Vision–ECCV 2016 Workshops: Amster- BIBLIOGRAPHY 175 dam, The Netherlands, October 8-10 and 15-16, 2016, Proceedings, Part III

2016
[217]

Rethinking the inception architecture for computer vision

Christian Szegedy et al. “Rethinking the inception architecture for computer vision”. In:Proceedings of the IEEE conference on computer vision and pattern recognition. 2016, pp. 2818–2826

2016
[218]

Deep Residual Learning for Small-footprint Keyword Spotting

Raphael Tang and Jimmy Lin. Deep Residual Learning for Small-footprint Keyword Spotting. Tech. rep. 2018. arXiv:1710.10361v2

2018 arXiv
[219]

Sound spectrographic cry analysis of pain cry in prematures

Carl-Johan Thodén, Anna-Liisa Järvenpää, and Katarina Michelsson. “Sound spectrographic cry analysis of pain cry in prematures”. In: Infant Crying. Springer, 1985, pp. 105–117

1985
[220]

Principles of Voice Production

Ingo R. Titze. “Principles of Voice Production”. In: (1994)

1994
[221]

Compressing recur- rent neural network with tensor train

Andros Tjandra, Sakriani Sakti, and Satoshi Nakamura. “Compressing recur- rent neural network with tensor train”. In:International Joint Conference on Neural Networks (IJCNN). IEEE. 2017, pp. 4451–4458

2017
[222]

2016, pp

Springer. 2016, pp. 443–450

2016
[223]

Unbiased look at dataset bias

Antonio Torralba and Alexei A. Efros. “Unbiased look at dataset bias”. In: CVPR 2011. 2011, pp. 1521–1528.doi: 10.1109/CVPR.2011.5995347

2011
[224]

The role of task and acous- tic similarity in audio transfer learning: insights from the speech emotion recog- nition case

Andreas Triantafyllopoulos and Björn W Schuller. “The role of task and acous- tic similarity in audio transfer learning: insights from the speech emotion recog- nition case”. In:ICASSP. IEEE. 2021

2021
[225]

CRY SOUNDS OF THE NEWBORN INFANT 1, 2

Henry Mayer Truby and John Lind. “CRY SOUNDS OF THE NEWBORN INFANT 1, 2”. In:Acta paediatrica54 (1965), pp. 8–59

1965
[226]

CryCeleb2023 Diary Blog

Dien-Hoa Truong. CryCeleb2023 Diary Blog. https://dienhoa.github.io/ dhblog/posts/cryceleb.html. Accessed: 2023-09-13. BIBLIOGRAPHY 176

2023
[227]

Some Mathematical Notes on Three-Mode Factor Anal- ysis

Ledyard R. Tucker. “Some Mathematical Notes on Three-Mode Factor Anal- ysis”. In:Psychometrika 31.3 (1966), pp. 279–311

1966
[228]

Unbiased look at dataset bias

Antonio Torralba and Alexei A Efros. “Unbiased look at dataset bias”. In: CVPR 2011. IEEE. 2011, pp. 1521–1528

2011
[229]

Adversarial discriminative domain adaptation

Eric Tzeng et al. “Adversarial discriminative domain adaptation”. In:Proceed- ings of the IEEE conference on computer vision and pattern recognition. 2017, pp. 7167–7176

2017
[230]

Deep domain confusion: Maximizing for domain invariance

Eric Tzeng et al. “Deep domain confusion: Maximizing for domain invariance”. In: arXiv preprint arXiv:1412.3474(2014)

2014 arXiv
[231]

Eric Tzeng et al.Simultaneous Deep Transfer Across Domains and Tasks. 2015. arXiv: 1510.02192 [cs.CV]

2015 arXiv
[232]

The Ubenwa CryCeleb Dataset

Ubenwa. The Ubenwa CryCeleb Dataset. 2023. doi: 10.57967/hf/1014. url: https://huggingface.co/datasets/Ubenwa/CryCeleb2023

2023 doi
[233]

Trans- fer learning by supervised pre-training for audio-based music classification

Aäron Van Den Oord, Sander Dieleman, and Benjamin Schrauwen. “Trans- fer learning by supervised pre-training for audio-based music classification”. In: Conference of the International Society for Music Information Retrieval (ISMIR 2014). 2014

2014
[234]

Musical genre classification of audio sig- nals

George Tzanetakis and Perry Cook. “Musical genre classification of audio sig- nals”.In: IEEE Transactions on Speech and Audio Processing.Vol.10.5.IEEE, 2002, pp. 293–302

2002
[235]

Efficient classical simulation of slightly entangled quantum com- putations

Guifré Vidal. “Efficient classical simulation of slightly entangled quantum com- putations”. In:Physical Review Letters91.14 (2003), p. 147902. BIBLIOGRAPHY 177

2003
[236]

Generalized end-to-end loss for speaker verification

Li Wan et al. “Generalized end-to-end loss for speaker verification”. In:2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2018, pp. 4879–4883

2018
[237]

Towards learning universal audio representations

Luyu Wang et al. “Towards learning universal audio representations”. In: ICASSP. IEEE. 2022

2022
[238]

Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition

Pete Warden. Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition. 2018. arXiv:1804.03209 [cs.CL]

2018 arXiv
[239]

Spectrographic analysis of pain cry in hyperbilirubine- mia

O Wasz-Höckert et al. “Spectrographic analysis of pain cry in hyperbilirubine- mia”. In:Neonatology 17.3-4 (1971), pp. 260–271

1971
[240]

Uni- versity of Edinburgh

Christophe Veaux, Junichi Yamagishi, and Kirsten MacDonald.CSTR VCTK Corpus: English Multi-speaker Corpus for CSTR Voice Cloning Toolkit. Uni- versity of Edinburgh. The Centre for Speech Technology Research (CSTR), 2017

2017
[241]

Twenty-five years of Scandinavian cry research

Ole Wasz-Höckert, Katarina Michelsson, and John Lind. “Twenty-five years of Scandinavian cry research”. In:Infant crying: Theoretical and research per- spectives (1985), pp. 83–104

1985
[242]

Lavenham, UK: Spastics International Medical Publications, 1968

Ole Wasz-Höckert et al.The Infant Cry: A Spectrographic and Auditory Anal- ysis. Lavenham, UK: Spastics International Medical Publications, 1968

1968
[243]

A Learning Algorithm for Continu- ally Running Fully Recurrent Neural Networks

Ronald J. Williams and David Zipser. “A Learning Algorithm for Continu- ally Running Fully Recurrent Neural Networks”. In:Neural Computation 1.2 (1989), pp. 270–280

1989
[244]

Self-training with Noisy Student Improves ImageNet Classi- fication

Qizhe Xie et al. “Self-training with Noisy Student Improves ImageNet Classi- fication”. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2020, pp. 10687–10698

2020
[245]

Utterance-level aggregation for speaker recognition in the wild

Weidi Xie et al. “Utterance-level aggregation for speaker recognition in the wild”. In: ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2019, pp. 5791–5795. BIBLIOGRAPHY 178

2019
[246]

The identification of some specific meanings in infant vocalization

O Wasz-Höckert et al. “The identification of some specific meanings in infant vocalization”. In:Experientia 20 (1964)

1964
[247]

Improvingspeechenhancementthroughfine-grainedspeech characteristics

MuqiaoYangetal.“Improvingspeechenhancementthroughfine-grainedspeech characteristics”. In:arXiv preprint arXiv:2207.00237(2022)

2022 arXiv
[248]

SUPERB:Speechprocessinguniversalperformancebench- mark

Shu-wenYangetal.“SUPERB:Speechprocessinguniversalperformancebench- mark”. In:INTERSPEECH (2021)

2021
[249]

Tensor-train recurrent neural networks for video classification

Yinchong Yang, Denis Krompass, and Volker Tresp. “Tensor-train recurrent neural networks for video classification”. In:arXiv preprint arXiv:1707.01786 (2017)

2017 arXiv
[250]

Compressing Recurrent Neural Networks Using Hierarchical Tucker Tensor Decomposition

Miao Yin et al. “Compressing Recurrent Neural Networks Using Hierarchical Tucker Tensor Decomposition”. In:arXiv preprint arXiv:2005.04366(2020)

2020 arXiv
[251]

How transferable are features in deep neural networks? Tech

Jason Yosinski et al. How transferable are features in deep neural networks? Tech. rep.url: https://arxiv.org/pdf/1411.1792.pdf

arXiv
[252]

Larger norm more transferable: An adaptive feature norm ap- proachforunsuperviseddomainadaptation

Ruijia Xu et al. “Larger norm more transferable: An adaptive feature norm ap- proachforunsuperviseddomainadaptation”.In: Proceedings of the IEEE/CVF international conference on computer vision. 2019, pp. 1426–1435

2019
[253]

Long-term Forecasting Using Tensor Train RNN’s

Rose Yu et al. “Long-term Forecasting Using Tensor Train RNN’s”. In: 2017

2017
[254]

Visualizing and understanding convo- lutional networks

Matthew D Zeiler and Rob Fergus. “Visualizing and understanding convo- lutional networks”. In: Computer Vision–ECCV 2014: 13th European Con- ference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part I 13. Springer. 2014, pp. 818–833

2014
[255]

Analysis of infant crying

PHILIP SANFORD Zeskind and BM Lester. “Analysis of infant crying”. In: Biobehavioral assessment of the infant(2001), pp. 149–166

2001
[256]

Lq-nets: Learned quantization for highly accurate and compact deep neural networks

Dongqing Zhang et al. “Lq-nets: Learned quantization for highly accurate and compact deep neural networks”. In:Proceedings of the European conference on computer vision (ECCV). 2018, pp. 365–382. BIBLIOGRAPHY 179

2018
[257]

Importance Weighted Adversarial Nets for Partial Domain Adaptation

Jing Zhang et al. Importance Weighted Adversarial Nets for Partial Domain Adaptation. 2018. arXiv:1803.09210 [cs.CV]

2018 arXiv
[258]

How transferable are features in deep neural networks?

Jason Yosinski et al. “How transferable are features in deep neural networks?” In: Advances in neural information processing systems. 2014, pp. 3320–3328

2014
[259]

Training multi-task adversarial network for extracting noise-robust speaker embedding

Jianfeng Zhou et al. “Training multi-task adversarial network for extracting noise-robust speaker embedding”. In:ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2019, pp. 6196–6200

2019
[264]

Domain-Symmetric Networks for Adversarial Domain Adap- tation

YabinZhangetal. Domain-Symmetric Networks for Adversarial Domain Adap- tation. 2019. arXiv:1904.04663 [cs.CV]

2019 arXiv
[335]

isbn: 978-0-7695-3441-1. doi: 10 . 1109 / MICAI . 2008 . 73. url: http : //ieeexplore.ieee.org/document/4682484/

2008
[377]

issn: 0018-9278. doi: 10 . 1109 / TAU . 1972 . 1162410. url: http : / / ieeexplore.ieee.org/document/1162410/

1972
[512]

doi: 10.1109/IEMBS.2006.259802

isbn: 1-4244-0032-5. doi: 10.1109/IEMBS.2006.259802 . url: https: //ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4461798

2006
[1359]

url: http://ieeexplore.ieee.org/ document/5288526/

doi: 10.1109/TKDE.2009.191 . url: http://ieeexplore.ieee.org/ document/5288526/. BIBLIOGRAPHY 170

2009
[2021]

arXiv: 2106.04624 [eess.AS]

arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

A fully automated approach for baby cry signal segmentation and boundary detection of expi- ratory and inspiratory episodes

Lina Abou-Abbas, Chakib Tadj, and Hesam Alaie Fersaie. “A fully automated approach for baby cry signal segmentation and boundary detection of expi- ratory and inspiratory episodes”. In:The Journal of the Acoustical Society of America 142.3 (2017), pp. 1318–1331

2017

[2] [2]

VCMNet: Weakly supervised learning for automatic infant vocalisation maturity analysis

Najla Al Futaisi et al. “VCMNet: Weakly supervised learning for automatic infant vocalisation maturity analysis”. In:International Conference on Multi- modal Interaction. 2019

2019

[3] [3]

Tensor Decompositions for Learning Latent VariableModels

Animashree Anandkumar et al. “Tensor Decompositions for Learning Latent VariableModels”.In: Journal of Machine Learning Research15.1(2014),pp.2773– 2832

2014

[4] [4]

Alzheimer’s disease: Communication patternsinfamilialandsporadicforms

A. Appell J. Kertesz and M. Fisman. “Alzheimer’s disease: Communication patternsinfamilialandsporadicforms”.In: Archives of Neurology39.10(1982), pp. 649–652

1982

[5] [5]

On the differences between song and speech emotion recognition: Effect of feature sets, feature types, and classi- fiers

Bagus Tris Atmaja and Masato Akagi. “On the differences between song and speech emotion recognition: Effect of feature sets, feature types, and classi- fiers”. In: 2020 ieee region 10 conference (TENCON). IEEE. 2020, pp. 968– 972

2020

[6] [6]

Do deep nets really need to be deep?

Jimmy Ba and Rich Caruana. “Do deep nets really need to be deep?” In: Advances in neural information processing systems. 2014, pp. 2654–2662. 150 BIBLIOGRAPHY 151

2014

[7] [7]

A theory of learning from different domains

Shai Ben-David et al. “A theory of learning from different domains”. In:Ma- chine learning79 (2010), pp. 151–175

2010

[8] [8]

Analysis of Representations for Domain Adaptation

Shai Ben-David et al. “Analysis of Representations for Domain Adaptation”. In: Advances in Neural Information Processing Systems (NIPS). Curran Asso- ciates, Inc., 2007, pp. 137–144

2007