Pith. sign in

REVIEW 4 major objections 6 minor 41 references

Deep Learning-Based Feature Fusion for Emotion Analysis and Suicide Risk Differentiation in Chinese Psychological Support Hotlines

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A fusion of Wav2Vec2, pitch, and MFCC features with attention mechanisms improves negative-emotion detection in Chinese hotline calls, reaching an F1-score of 79.13%.

desk verdict Competent incremental feature-fusion paper with an honest null clinical result, but the claimed gains rest on single-run point estimates and the title oversells the evidence; worth refereeing with requests for significance tests and metric details. read the letter →

arxiv 2501.08696 v1 pith:F2B4CQIG submitted 2025-01-15 cs.CL

Han Wang , Jianqiang Li , Qing Zhao , Zhonglong Chen , Changwei Song , Jing Tang , Yuning Huang , Wei Zhai
show 2 more authors
Yongsheng Tong Guanghui Fu
This is my paper · ORCID
classification cs.CL
keywords speechemotionrecognitionfeaturefusionpsychologicalsupporthotlinesuicideriskassessmentWav2Vec2pitchfeaturesMFCCcross-attention
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that fusing three complementary acoustic feature types—self-supervised Wav2Vec2 representations, pitch (tone) features, and MFCCs—through cross-attention and self-attention improves speech emotion recognition over any single feature or prior fusion, especially in tonal languages like Chinese. On 20,254 audio segments from a major Chinese psychological support hotline, the fused model reaches an F1-score of 79.13% for binary negative/non-negative emotion classification, outperforming seven comparison models. The same approach also edges past the previous state of the art on the Vietnamese ViSEC dataset (73.30% vs. 72.72% accuracy), supporting the claim that tonal-feature fusion generalizes across tonal languages. Applied to 46 hotline callers, the model finds that people with suicidal behavior show more negative speech segments and a higher emotional change rate than non-suicidal callers, though the differences are not statistically significant. The paper argues these dynamic emotional-fluctuation measures could become novel features for suicide risk assessment, complementing static psychological scale scores.

What carries the argument

The load-bearing mechanism is a two-stage attention fusion. First, pitch features are projected through a non-pretrained Wav2Vec encoder to match the dimensionality of Wav2Vec2 embeddings, and the two are fused bidirectionally with cross-attention (each feature attends to the other). In parallel, 13 MFCC coefficients plus their first and second derivatives (39-dimensional) are encoded by a BiLSTM and average-pooled. The pitch-enhanced Wav2Vec feature and the MFCC feature are concatenated, then passed through a self-attention layer that builds connections among the fused features before a dropout-regularized classifier outputs the emotion probability. This design is what lets the model keep both high-level contextual representations and fine-grained tonal and spectral cues.

What would settle it

Re-running the proposed model on the same private dataset with a speaker-disjoint split constructed from scratch—ensuring no segment of a training subject is in the test set—would settle the central claim; if the F1-score falls back toward the single-Wav2Vec2 baseline of 76.96% rather than staying near 79.13%, the reported gain is likely an artifact of speaker leakage or label noise. Measuring inter-annotator agreement on a random sample of segments would test label reliability directly.

Watch

Extended reading notes

Core claim

The central claim is that a joint-attention feature fusion of Wav2Vec2 deep features, pitch features, and MFCCs yields the best-performing speech emotion recognition model for Chinese psychological support hotline data. The authors report an F1-score of 79.13% for negative emotion classification, with recall of 87.75%, beating Wav2Vec2, HuBERT, several Whisper variants, and the prior pitch-fusion method [19]. On the public ViSEC Vietnamese dataset, the model reaches 73.30% weighted and unweighted accuracy, above the previous state of the art's 72.72%. The paper further claims that when applied to hotline conversations, the model's outputs—counts of negative speech segments and an emotion change rate—show elevated emotional instability in the suicidal group, and suggests these metrics could inform suicide risk assessment, while acknowledging that the group differences were not statistically significant.

Load-bearing premise

The 79.13% F1-score assumes the three experts' negative/non-negative labels are consistent and that the patient-level 4:1 split prevents any speaker's audio from appearing in both training and test, since the paper reports no inter-annotator agreement and inherits the split from earlier work without a leakage check.

Editorial extensions

If this is right

  • If the central claim holds, hotline systems can use this fusion method to flag negative emotional segments in near real time, giving counselors a quantitative signal to prioritize during a call.
  • The ViSEC result suggests the tonal-feature fusion recipe transfers to other tonal languages, so the same architecture could be applied to Vietnamese, Thai, or Cantonese hotlines with language-matched Wav2Vec2 backbones.
  • Emotional change rate and negative-segment count become computable, dynamic descriptors of a caller's session, complementing one-time scale scores with a trajectory measure for risk research.
  • The non-significant group differences imply that distinguishing suicidal from non-suicidal callers will require larger cohorts or finer-grained labels than binary negative/non-negative emotion.

Reading between the lines

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

  • The emotion change rate is essentially a temporal variability statistic; the same formula could quantify emotional instability in other modalities, such as text-based chat counseling or longitudinal self-report mood data, without retraining the speech model.
  • Because the pitch pathway's encoder is randomly initialized, a dedicated pretrained tonal-feature encoder might improve the fusion further; the paper does not test this.
  • The reported F1 rests on a patient-level split inherited from prior work, so a speaker-disjoint replication with inter-annotator agreement would be the natural stress test of the 79.13% figure.
  • If the model's per-segment emotion probabilities were calibrated, the emotion change rate could be turned into an alerting threshold for counselors during the first 30-minute assessment stage.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. This paper proposes a speech emotion recognition (SER) model for Chinese psychological support hotlines. The model fuses Wav2Vec2 deep features, pitch features, and MFCC features through cross-attention and self-attention mechanisms. The authors report an F1-score of 79.13% for binary negative-emotion classification on a private hotline dataset, claiming it outperforms seven comparison models, and report 73.30% unweighted/weighted accuracy on the public ViSEC tonal-language dataset, claiming it outperforms prior state-of-the-art. The paper also applies the model to compare emotional fluctuation metrics (NSS and ECR) between 22 subjects with suicidal behavior and 24 without; the differences are not statistically significant, a limitation the paper explicitly acknowledges.

Significance. If the classification gains are real and reproducible, the contribution is a modest but useful SER feature-fusion recipe for tonal languages, with the rare addition of an external public dataset and an honest null clinical analysis. Strengths include the public source code, patient-level data splitting, evaluation on a held-out subject set, and transparent reporting that the clinical group differences are not statistically significant. The central limitation is that the claimed superiority over baselines currently rests on single point estimates with no uncertainty quantification, and the reported ViSEC metrics contain a suspicious pattern that needs verification before the central claim can be accepted.

major comments (4)
  1. [§V-B, Tables V and VI] No significance tests or variance estimates are provided for any classification comparison. The reported gains over the closest baselines are 1.32 F1 points on a 4,051-segment test set from only 21 subjects (Table V) and 0.58 accuracy points on 528 ViSEC samples (Table VI). Given the stochasticity of deep learning and the effective sample size, these differences are well within plausible run-to-run noise. Please report multiple random seeds, bootstrap confidence intervals at the subject level, and McNemar tests (or a suitable paired test) for the proposed model versus each baseline.
  2. [§V-B2, Table VI] The reported values are suspicious: the proposed model achieves UA=WA=73.30% and Thanh et al. achieves UA=WA=72.72%, yet the ViSEC test partition is imbalanced (Table II: per-class counts 123/150/108/147). For unweighted and weighted accuracy to coincide to two decimals for two different models is highly implausible unless the metrics were computed incorrectly or a balanced subset was used. Please provide the exact formulas, per-class confusion matrices, and corrected UA/WA values if necessary.
  3. [§IV-A0a, Table I, §V-B1] The evaluation metrics are computed at the segment level, but the test set contains only 21 subjects, and segments from the same subject are correlated. The current F1 of 79.13% may therefore overstate the model's generalizability to new callers. Please add a subject-level evaluation (e.g., per-subject F1 with confidence intervals) and explicitly describe the patient-level split procedure, including a check that no test subject appears in the training set; deferring to Chen et al. [10] is insufficient for a standalone reproducibility claim.
  4. [§IV-A0a] The private dataset labels are the supervision signal for the central classification claim, but no inter-annotator agreement measure is reported for the negative/non-negative annotations. Even though the labels are said to follow Chen et al. [10], the paper should report or cite a kappa value, since label noise directly affects the F1 estimate and the downstream clinical analysis.
minor comments (6)
  1. [Table II] The Total row for the ViSEC training split is incorrect: the sum of the training column is 4,224, not 3,030 (the validation and test columns do sum to 528 each). Please correct this inconsistency.
  2. [§III-C] The text refers to the "Thanh activation function"; this appears to be a typo for "Tanh activation function". Please clarify.
  3. [Index Terms] The index terms contain a malformed entry, "Feature -, usion", which should be corrected to "Feature fusion".
  4. [§V-B1, Table V] The proposed model achieves lower accuracy (75.19%) than Wav2Vec2 (76.97%) while obtaining a higher F1-score due to much higher recall (87.75% vs. 76.96%). The text should explicitly discuss this accuracy/F1 trade-off and explain why F1 is the appropriate primary metric for this application, rather than simply claiming overall superiority.
  5. [Equation (7)] There is a typo in the word "pervious"; it should be "previous".
  6. [Abstract and Discussion] The claim that emotional fluctuation intensity and frequency "could serve as novel features" for suicide risk assessment is speculative given that the reported differences are not statistically significant. The authors acknowledge this in the results, but the abstract and discussion should be tempered to avoid overstating the clinical implication.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: central classification claim is validated on held-out subjects and an external dataset; only minor non-load-bearing self-citations appear.

full rationale

The paper's core claim—that fusing Wav2Vec2, pitch, and MFCC features with cross-/self-attention improves emotion classification—is not circular. The model is trained on 84 subjects and tested on 21 held-out subjects (Table I), with an additional external benchmark on the ViSEC dataset where it is compared against Wav2Vec2 and Thanh et al. (Table VI). These evaluations provide independent evidence that does not reduce by construction to the fitted parameters. The clinical emotion-trend analysis applies the trained model to 46 subjects and computes NNS and ECR (Eq. 7); these metrics are hypotheses about emotional dynamics, not quantities fitted to suicide labels, and the paper honestly reports non-significant p-values. The only self-citations are to Chen et al. [10] for the annotation/split protocol and Song et al. [12] for hotline details; these are dataset-construction references with overlapping authors, but they are not load-bearing for the architectural or comparative conclusions, and the external ViSEC result plus held-out split give the central claim independent content. Concerns about missing significance tests or point estimates are correctness/statistical-robustness risks, not circularity. Score 2 reflects the minor self-citations only.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central claim rests on standard pre-trained models, hand-picked hyperparameters, and assumptions about annotation quality, data split integrity, and cross-lingual transfer. No new physical or conceptual entities are introduced; the emotion change rate is a defined metric, not an entity.

free parameters (5)
  • Audio clip length = 10 seconds
    All audio clips are padded or truncated to 10 seconds to fit GPU memory (Section IV-B); this discards context from longer hotline segments and is a hand-chosen preprocessing parameter.
  • MFCC feature dimension = 39
    13 MFCC coefficients plus first and second derivatives (Section III-A); standard but a hand-selected representation.
  • Learning rate = 3e-5
    Fixed for all training runs (Section IV-D); not tuned on validation.
  • Training epochs = 30
    All models trained for exactly 30 epochs (Section IV-D); no early stopping or validation-based selection reported.
  • Batch size = 8
    Chosen for 24GB GPU memory (Section IV-D); not swept.
assumptions (6)
  • domain assumption The expert annotations of negative and non-negative emotions on the private dataset are accurate and consistent.
    Section IV-A0a states segments were 'annotated by three experts', but no inter-annotator agreement is reported, so label noise is unmeasured.
  • domain assumption The patient-level 4:1 split prevents speaker leakage between training and test sets.
    Section IV-A0a says the split was done 'at the patient level' and follows Chen et al. [10], but no deduplication check or subject-level statistics are provided.
  • domain assumption Pitch and MFCC features carry emotional information complementary to Wav2Vec2 features for tonal languages.
    This is the premise of the fusion design (Section III-A) and is supported only by the paper's own ablation on one dataset.
  • domain assumption The ViSEC dataset is a valid external benchmark for a model designed for Chinese speech.
    Section IV-A0b justifies the choice by shared tonality, but Chinese and Vietnamese are different languages with different acoustic distributions.
  • domain assumption The first 30 minutes of a hotline call correspond to the suicide risk assessment stage.
    Section IV-E relies on the hotline workflow in Tong et al. [33], [34], but the specific 30-minute boundary is not justified with data.
  • domain assumption The pre-trained Wav2Vec2 models used (Chinese and Vietnamese) are suitable feature extractors for the target audio.
    Section IV-D selects models used in prior work [10], [19]; no verification of embedding quality on hotline audio is provided beyond final performance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Learning-Based Feature Fusion for Emotion Analysis and Suicide Risk Differentiation in Chinese Psychological Support Hotlines." pith.science (2026). https://pith.science/paper/F2B4CQIG

@misc{pith2026250108696,
  author       = {Pith},
  title        = {Pith review of: Deep Learning-Based Feature Fusion for Emotion Analysis and Suicide Risk Differentiation in Chinese Psychological Support Hotlines},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F2B4CQIG}},
  note         = {Machine review of arXiv:2501.08696}
}
read the original abstract

Mental health is a critical global public health issue, and psychological support hotlines play a pivotal role in providing mental health assistance and identifying suicide risks at an early stage. However, the emotional expressions conveyed during these calls remain underexplored in current research. This study introduces a method that combines pitch acoustic features with deep learning-based features to analyze and understand emotions expressed during hotline interactions. Using data from China's largest psychological support hotline, our method achieved an F1-score of 79.13% for negative binary emotion classification.Additionally, the proposed approach was validated on an open dataset for multi-class emotion classification,where it demonstrated better performance compared to the state-of-the-art methods. To explore its clinical relevance, we applied the model to analysis the frequency of negative emotions and the rate of emotional change in the conversation, comparing 46 subjects with suicidal behavior to those without. While the suicidal group exhibited more frequent emotional changes than the non-suicidal group, the difference was not statistically significant.Importantly, our findings suggest that emotional fluctuation intensity and frequency could serve as novel features for psychological assessment scales and suicide risk prediction.The proposed method provides valuable insights into emotional dynamics and has the potential to advance early intervention and improve suicide prevention strategies through integration with clinical tools and assessments The source code is publicly available at https://github.com/Sco-field/Speechemotionrecognition/tree/main.

Figures

Figures reproduced from arXiv: 2501.08696 by the authors.

Figure 1
Figure 1. The model architecture of the proposed method. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Emotional trends (from Negative to Non-negative) of callers during the full conversation. The “Non-suicide” group [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 28 canonical work pages

  1. [10]

    Fine-grained Speech Sentiment Analysis in Chinese Psychological Support Hotlines Based on Large-scale Pre-trained Model

    Z. Chen, C. Song, Y . Chen, J. Li, G. Fu, Y . Tong, and Q. Zhao, “Fine-grained speech sentiment analysis in chinese psychological sup- port hotlines based on large-scale pre-trained model,” arXiv preprint arXiv:2405.04128, 2024

    work page Pith review arXiv 2024

  2. [1]

    The integrated motivational– volitional model of suicidal behaviour,

    R. C. O’Connor and O. J. Kirtley, “The integrated motivational– volitional model of suicidal behaviour,” Philosophical Transactions of the Royal Society B: Biological Sciences , vol. 373, no. 1754, p. 20170268, 2018

    work page 2018

  3. [2]

    Suicide research in mainland china, hong kong, and macau over three decades,

    Q. Cheng, X. Zhang, C. Lui, and P. S. Yip, “Suicide research in mainland china, hong kong, and macau over three decades,” Crisis, 2020

    work page 2020

  4. [3]

    The utility of artificial intelligence in suicide risk prediction and the management of suicidal behaviors,

    T. M. Fonseka, V . Bhat, and S. H. Kennedy, “The utility of artificial intelligence in suicide risk prediction and the management of suicidal behaviors,” Australian & New Zealand Journal of Psychiatry , vol. 53, no. 10, pp. 954–964, 2019

    work page 2019

  5. [4]

    Suicide-related calls to a national crisis chat hotline service during the covid-19 pandemic and lockdown,

    G. Zalsman, Y . Levy, E. Sommerfeld, A. Segal, D. Assa, L. Ben- Dayan, A. Valevski, and J. J. Mann, “Suicide-related calls to a national crisis chat hotline service during the covid-19 pandemic and lockdown,” Journal of psychiatric research , vol. 139, pp. 193–196, 2021

    work page 2021

  6. [5]

    Follow-up with callers to the national suicide prevention lifeline: Evaluation of callers’ perceptions of care,

    M. S. Gould, A. M. Lake, H. Galfalvy, M. Kleinman, J. L. Munfakh, J. Wright, and R. McKeon, “Follow-up with callers to the national suicide prevention lifeline: Evaluation of callers’ perceptions of care,” Suicide and Life-Threatening Behavior , vol. 48, no. 1, pp. 75–86, 2018

    work page 2018

  7. [6]

    Helping callers to the national suicide prevention lifeline who are at imminent risk of suicide: Evalua- tion of caller risk profiles and interventions implemented,

    M. S. Gould, A. M. Lake, J. L. Munfakh, H. Galfalvy, M. Kleinman, C. Williams, A. Glass, and R. McKeon, “Helping callers to the national suicide prevention lifeline who are at imminent risk of suicide: Evalua- tion of caller risk profiles and interventions implemented,” Suicide and Life-Threatening Behavior, vol. 46, no. 2, pp. 172–190, 2016

    work page 2016

  8. [7]

    Biologically inspired speech emo- tion recognition,

    R. Lotfidereshgi and P. Gournay, “Biologically inspired speech emo- tion recognition,” in 2017 IEEE international conference on acoustics, speech and signal processing (ICASSP) . IEEE, 2017, pp. 5135–5139

    work page 2017

Show all 41 references
  1. [8]

    An evaluation of suicide prevention hotline results in taiwan: caller profiles and the effect on emotional distress and suicide risk,

    F. F.-T. Shaw and W.-H. Chiang, “An evaluation of suicide prevention hotline results in taiwan: caller profiles and the effect on emotional distress and suicide risk,” Journal of affective disorders , vol. 244, pp. 16–20, 2019

    2019

  2. [9]

    Unlocking the emotional states of high-risk suicide callers through speech analysis,

    A. Nfissi, W. Bouachir, N. Bouguila, and B. Mishara, “Unlocking the emotional states of high-risk suicide callers through speech analysis,” in 2024 IEEE 18th International Conference on Semantic Computing (ICSC). IEEE, 2024, pp. 33–40

    2024

  3. [11]

    Speech emotion recognition approaches: A systematic review,

    A. Hashem, M. Arif, and M. Alghamdi, “Speech emotion recognition approaches: A systematic review,” Speech Communication , p. 102974, 2023

    2023

  4. [12]

    An exploratory deep learning approach for predicting subsequent suicidal acts in chinese psychological support hotlines,

    C. Song, Q. Zhao, J. Li, Y . Chen, Y . Tong, and G. Fu, “An exploratory deep learning approach for predicting subsequent suicidal acts in chinese psychological support hotlines,” arXiv preprint arXiv:2408.16463, 2024

    2024 arXiv

  5. [13]

    Emotion recognition from natural phone conversations in individuals with and without recent suicidal ideation,

    J. Gideon, H. T. Schatten, M. G. McInnis, and E. M. Provost, “Emotion recognition from natural phone conversations in individuals with and without recent suicidal ideation,” in Interspeech, 2019

    2019

  6. [14]

    The geneva minimalistic acoustic parameter set (gemaps) for voice research and affective computing,

    F. Eyben, K. R. Scherer, B. W. Schuller, J. Sundberg, E. Andr´e, C. Busso, L. Y . Devillers, J. Epps, P. Laukka, S. S. Narayanan et al., “The geneva minimalistic acoustic parameter set (gemaps) for voice research and affective computing,” IEEE transactions on affective computi...

    2015

  7. [15]

    Spontaneous speech-based suicide risk detection using whisper and large language models,

    Z. Cui, C. Lei, W. Wu, Y . Duan, D. Qu, J. Wu, R. Chen, and C. Zhang, “Spontaneous speech-based suicide risk detection using whisper and large language models,” arXiv preprint arXiv:2406.03882 , 2024. 10

    2024 arXiv

  8. [16]

    Robust speech recognition via large-scale weak supervi- sion,

    A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust speech recognition via large-scale weak supervi- sion,” in International conference on machine learning . PMLR, 2023, pp. 28 492–28 518

    2023

  9. [17]

    Speech emotion recognition with co-attention based multi-level acoustic information,

    H. Zou, Y . Si, C. Chen, D. Rajan, and E. S. Chng, “Speech emotion recognition with co-attention based multi-level acoustic information,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2022, pp. 7367–7371

    2022

  10. [18]

    Dual-tbnet: Improving the robustness of speech features via dual-transformer-bilstm for speech emotion recognition,

    Z. Liu, X. Kang, and F. Ren, “Dual-tbnet: Improving the robustness of speech features via dual-transformer-bilstm for speech emotion recognition,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 31, pp. 2193–2203, 2023

    2023

  11. [19]

    A robust pitch-fusion model for speech emotion recognition in tonal languages,

    P. V . Thanh, N. T. T. Huyen, P. N. Quan, and N. T. T. Trang, “A robust pitch-fusion model for speech emotion recognition in tonal languages,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 12 386– 12 390

    2024

  12. [20]

    Emotion neural transducer for fine-grained speech emotion recognition,

    S. Shen, Y . Gao, F. Liu, H. Wang, and A. Zhou, “Emotion neural transducer for fine-grained speech emotion recognition,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 10 111–10 115

    2024

  13. [21]

    Deep learning based speech emotion recognition for parkinson patient,

    H. Khan, M. Ullah, F. Al-Machot, F. A. Cheikh, and M. Sajjad, “Deep learning based speech emotion recognition for parkinson patient,” Electronic Imaging, vol. 35, pp. 298–1, 2023

    2023

  14. [22]

    Emotion recognition from speech using mfcc and dwt for security system,

    S. T. Saste and S. Jagdale, “Emotion recognition from speech using mfcc and dwt for security system,” in 2017 international conference of electronics, communication and aerospace technology (ICECA) , vol. 1. IEEE, 2017, pp. 701–704

    2017

  15. [23]

    Investigating trans- former encoders and fusion strategies for speech emotion recognition in emergency call center conversations

    T. Deschamps-Berger, L. Lamel, and L. Devillers, “Investigating trans- former encoders and fusion strategies for speech emotion recognition in emergency call center conversations.” in Companion Publication of the 2022 International Conference on Multimodal Interaction , 2022, ...

    2022

  16. [24]

    Emotion recognition from speech: a review,

    S. G. Koolagudi and K. S. Rao, “Emotion recognition from speech: a review,”International journal of speech technology, vol. 15, pp. 99–117, 2012

    2012

  17. [25]

    Speech emotion recognition via multi-level attention network,

    K. Liu, D. Wang, D. Wu, Y . Liu, and J. Feng, “Speech emotion recognition via multi-level attention network,” IEEE Signal Processing Letters, vol. 29, pp. 2278–2282, 2022

    2022

  18. [26]

    Stutternet: Stuttering disfluencies detection in synthetic speech signals via mel frequency cepstral coefficients features using deep learning,

    M. Abubakar, M. Mujahid, K. Kanwal, S. Iqbal, N. Asghar, and A. Alaulamie, “Stutternet: Stuttering disfluencies detection in synthetic speech signals via mel frequency cepstral coefficients features using deep learning,” IEEE Access, 2024

    2024

  19. [27]

    Attention is all you need,

    A. Vaswani, “Attention is all you need,” Advances in Neural Information Processing Systems, 2017

    2017

  20. [28]

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

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

    2020

  21. [29]

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

    W.-N. Hsu, B. Bolte, Y .-H. H. Tsai, K. Lakhotia, R. Salakhutdinov, and A. Mohamed, “Hubert: Self-supervised speech representation learning by masked prediction of hidden units,” IEEE/ACM transactions on audio, speech, and language processing , vol. 29, pp. 3451–3460, 2021

    2021

  22. [30]

    Automatic speech recognition system for tonal languages: State-of-the-art survey,

    J. Kaur, A. Singh, and V . Kadyan, “Automatic speech recognition system for tonal languages: State-of-the-art survey,” Archives of Computational Methods in Engineering , vol. 28, pp. 1039–1068, 2021

    2021

  23. [31]

    Mispronun- ciation detection and diagnosis model for tonal language, applied to vietnamese,

    T. T. Huu, V . T. Pham, T. T. T. Nguyen, and T. L. Dao, “Mispronun- ciation detection and diagnosis model for tonal language, applied to vietnamese,” in Proc. INTERSPEECH 2023 , 2023, pp. 1014–1018

    2023

  24. [32]

    Multi-level speaker- independent emotion recognition using complex-mfcc and swin trans- former,

    M. Saadati, R. M. Toroghi, and H. Zareian, “Multi-level speaker- independent emotion recognition using complex-mfcc and swin trans- former,” in 2024 20th CSI International Symposium on Artificial Intel- ligence and Signal Processing (AISP) . IEEE, 2024, pp. 1–4

    2024

  25. [33]

    Prospective study of association of characteristics of hotline psychological intervention in 778 high-risk callers with subsequent suicidal act,

    Y . Tong, K. R. Conner, C. Wang, Y . Yin, L. Zhao, Y . Wang, and Y . Liu, “Prospective study of association of characteristics of hotline psychological intervention in 778 high-risk callers with subsequent suicidal act,” Australian & New Zealand Journal of Psychiatry , vol. 54...

    2020

  26. [34]

    Predictive value of suicidal risk assessment using data from china’s largest suicide prevention hotline,

    Y . Tong, Y . Yin, K. R. Conner, L. Zhao, Y . Wang, X. Wang, and Y . Conwell, “Predictive value of suicidal risk assessment using data from china’s largest suicide prevention hotline,”Journal of affective disorders, vol. 329, pp. 141–148, 2023

    2023

  27. [35]

    Fleurs: Few-shot learning evaluation of universal representations of speech,

    A. Conneau, M. Ma, S. Khanuja, Y . Zhang, V . Axelrod, S. Dalmia, J. Riesa, C. Rivera, and A. Bapna, “Fleurs: Few-shot learning evaluation of universal representations of speech,” 2022. [Online]. Available: https://arxiv.org/abs/2205.12446

    2022 arXiv

  28. [36]

    Com- mon voice: A massively-multilingual speech corpus,

    R. Ardila, M. Branson, K. Davis, M. Henretty, M. Kohler, J. Meyer, R. Morais, L. Saunders, F. M. Tyers, and G. Weber, “Com- mon voice: A massively-multilingual speech corpus,” arXiv preprint arXiv:1912.06670, 2019

    1912 arXiv

  29. [37]

    Css10: A collection of single speaker speech datasets for 10 languages,

    K. Park and T. Mulc, “Css10: A collection of single speaker speech datasets for 10 languages,” arXiv preprint arXiv:1903.11269 , 2019

    1903 arXiv

  30. [38]

    St-cmds-20170001 1, free st chinese mandarin corpus,

    “St-cmds-20170001 1, free st chinese mandarin corpus,” https://openslr. org/38, 2017, accessed: 2023-10-15

    2017

  31. [39]

    Wav2vec2-base-vi,

    N. Vulebinh, “Wav2vec2-base-vi,” https://huggingface.co/ nguyenvulebinh/wav2vec2-base-vi, 2023, accessed: 15 October 2023

    2023

  32. [40]

    librosa: Audio and music signal analysis in python

    B. McFee, C. Raffel, D. Liang, D. P. Ellis, M. McVicar, E. Battenberg, and O. Nieto, “librosa: Audio and music signal analysis in python.” in SciPy, 2015, pp. 18–24

    2015

  33. [41]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint, vol. 1412.6980, 2014

    2014 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.