Pith. sign in

REVIEW 4 major objections 6 minor 31 references

Speaker Anonymisation for Speech-based Suicide Risk Detection

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper shows that ensembling a prosody-preserving voice converter with a semantic-preserving synthesizer yields suicide-risk detection accuracy of 0.692, within one percentage point of original speech, while anonymising the speaker.

desk verdict Useful systematic comparison, but the privacy claim rests on an unspecified EER protocol and the utility claim lacks error bars. read the letter →

arxiv 2509.22148 v2 pith:OIIQ57KB submitted 2025-09-26 eess.AS cs.SD

Ziyun Cui , Sike Jia , Yang Lin , Yinan Duan , Diyang Qu , Runsen Chen , Chao Zhang , Chang Lei
show 1 more author
Wen Wu
This is my paper · ORCID
classification eess.AScs.SD
keywords speakeranonymisationsuicideriskdetectionadolescentspeechvoiceconversionsynthesisprivacyevaluationlargelanguagemodelensembleofmethods
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 asks whether adolescent speech can be anonymised without destroying the acoustic, prosodic, and semantic cues used to detect suicide risk, a privacy-utility trade-off that matters for a doubly vulnerable population. It systematically compares seven anonymisation pipelines—pitch shifting, McAdams formant modification, four neural voice-conversion systems, and two text-to-speech synthesizers—on Mandarin adolescent clinical recordings. The central claim is that no single method suffices, but combining RVC (which preserves prosodic and acoustic contours) with CosyVoice (which preserves semantic content) yields a suicide-risk detection accuracy of 0.692, a gap of only one percentage point to the 0.702 obtained on original speech, while raising speaker-verification equal error rate to roughly 0.5. In plain terms, the paper argues that privacy protection and clinical utility are not mutually exclusive if anonymisation choices are treated as complementary.

What carries the argument

The load-bearing mechanism is the complementarity of two anonymisation families, operationalised through the paper's five-metric evaluation framework (signal quality via SNR and UTMOS MOS, F0-contour deviation via L1 distance and Pearson correlation, semantic preservation via character error rate, emotional preservation via Emotion2Vec cosine similarity) and speaker discrimination via equal error rate. RVC, a retrieval-based neural voice converter, preserves F0 contours and acoustic detail; CosyVoice, a cascaded ASR-plus-TTS synthesizer, discards acoustics but keeps semantic content. Their predicted-probability ensemble is the central object that carries the argument. The detection backbone is Qwen2.5-Omni-7B, a speech large language model fine-tuned with weight-decomposed low-rank adaptation (DoRA, rank 32, alpha 64).

What would settle it

Run a realistic linkage attack: enroll a strong speaker-verification model on original speech and test it against the RVC+CosyVoice anonymised utterances. If the achieved EER under this protocol falls well below the reported ~0.5, or if the original-speech EER is far below 0.185 when measured properly, then the 'good anonymisation' claim fails even though detection accuracy may remain near 0.692.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the information lost by one anonymisation method is often retained by another, and that a simple probability-average ensemble of two complementary pipelines can recover most of the original speech's clinical signal. In the authors' evaluation, RVC retains F0-contour shape and acoustic detail better than any TTS method, while CosyVoice preserves semantic content (character error rate 0.024 versus 0.332 for SparkTTS) at the cost of prosodic and emotional information. Averaging the suicide-risk probabilities predicted by detectors trained on RVC-anonymised and CosyVoice-anonymised speech yields 0.692 accuracy, surpassing every individual system and coming within 0.01 of the 0.702 accuracy of original recordings, with a speaker-verification EER around 0.5 indicating substantial identity masking. The paper takes this as evidence that hybrid anonymisation strategies are viable for privacy-preserving clinical speech analytics.

Load-bearing premise

The load-bearing assumption is that the paper's equal-error-rate numbers really capture how hard it is to identify the original speaker from anonymised audio; but the verification model and attack protocol are not specified, and even the original recordings already show a 0.185 error rate, so the privacy guarantee could be much weaker against a stronger attacker.

Editorial extensions

If this is right

  • If the central claim holds, anonymised clinical speech can be safely shared for suicide-risk screening, since the 1% accuracy gap is small enough to preserve population-level utility.
  • Hybrid anonymisation—pairing an acoustic-preserving voice-conversion method with a semantic-preserving synthesis method—becomes a general template for balancing privacy and utility in other paralinguistic clinical tasks such as depression or voice-disorder screening.
  • The finding that speech enhancement before anonymisation does not consistently help (and can hurt) downstream accuracy implies that pipeline engineering should focus on compatibility between enhancement and anonymisation, not just signal-to-noise ratio.
  • The five-metric evaluation framework plus EER provides a reusable benchmark for future anonymisation methods targeting clinical speech.
  • Since different anonymisations degrade different cue channels, detector ensembles trained on multiple anonymised views may prove more robust than any single pipeline, a direct extension already demonstrated for RVC and CosyVoice.

Reading between the lines

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

  • Inference: the privacy guarantee is only as strong as the unspecified speaker-verification evaluation—original speech already has an EER of 0.185, so a stronger attacker or a more realistic linkage protocol could shrink the apparent anonymity margin; reporting the verification model and enrollment/test conditions would settle this.
  • Inference: the ensemble's success hints that suicide-risk information is distributed redundantly across semantic and prosodic channels, so a testable prediction is that ablating either channel will consistently lose accuracy, and adding a third complementary anonymisation could close the remaining 1% gap.
  • Inference: the same complementarity principle could transfer to other privacy-sensitive speech tasks, and the simplest extension would be to replace the suicide-risk classifier with a different paralinguistic head while keeping the anonymisation ensemble fixed.
  • Inference: calibrated score fusion instead of simple probability averaging might yield further gains, but this is untested in the paper.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper studies speaker anonymisation for speech-based suicide risk detection in adolescent speech. It evaluates a broad set of anonymisation methods (pitch modification, McAdams, SSL-SAS, FreeVC, SeedVC, RVC, and TTS-based SparkTTS/CosyVoice) under a five-metric framework (SNR, MOS, F0 deviation, semantic CER, emotion similarity, and speaker EER). The central claim is that a combination of CosyVoice and RVC, fused by averaging predicted probabilities, achieves suicide-risk detection accuracy of 0.692 against 0.702 for original speech, while still providing speaker anonymisation. The authors also test the effect of FRCRN speech enhancement prior to anonymisation and find no consistent benefit.

Significance. If the central claim holds, the paper makes a useful contribution by showing that anonymised speech can preserve enough clinical information for suicide risk detection while reducing speaker traceability in a vulnerable population. The study is the first systematic comparison of this kind for this task, and the detection evaluation is largely external: the detector is fine-tuned on original speech and evaluated on anonymised speech, with anonymisation systems pretrained off-the-shelf. The multi-metric evaluation framework and the exploration of complementary information across anonymisation methods are strengths. However, the headline 'comparable performance' and 'good anonymisation' claims rest on statistical and measurement gaps that need to be addressed before the findings can be considered robust.

major comments (4)
  1. [Section 4.2, Fig. 4] The claim of 'comparable performance' (0.692 vs 0.702, a 1% gap) is not supported by any measure of uncertainty. Accuracy is averaged over three random seeds on a test set of roughly 122 samples (1,223 participants, 8:1:1 split), so the standard error of the mean is large; a 0.01 difference is well within run-to-run variability. Please report per-run results, standard deviations or confidence intervals, and perform a paired significance test (e.g., McNemar or bootstrap) between the ensemble and original speech, and between the ensemble and RVC alone. Without this, the central 'comparable performance' conclusion is not established.
  2. [Section 2.3, Table 1] The EER values used to support the 'good anonymisation' half of the trade-off are not backed by any specification of the speaker verification protocol: the verification model (e.g., ECAPA-TDNN, WavLM-based), enrollment utterances, test segments, trial construction (same/different speakers, number of impostors), and attack scenario are all unspecified. This is particularly problematic because the original speech already has EER=0.185, which is far from the near-zero value expected for unmodified audio, indicating an unusually difficult or weak verification setup. Under a stronger attacker, the EER margins for RVC (0.510) and CosyVoice (0.497) could shrink substantially, undermining the 'protection of speaker identity' claim. Please describe the verification protocol in detail and, ideally, evaluate with a stronger or state-of-the-art verifier, reporting DET curves or EER with confidence intervals.
  3. [Section 4.2 and Table 1] The final combined CosyVoice+RVC system is never evaluated for speaker anonymisation. The paper reports EER only for individual methods, but the deployed composite system exposes both anonymised versions of each speaker. An attacker holding both outputs could link them through shared lexical content, timing, and residual prosodic cues, and the per-component EERs do not cover this union-of-outputs threat. Please compute an EER for the combined system, ideally under a protocol where enrollment and test come from the same composite output distribution, or at least clearly state that the privacy guarantee does not extend to the ensemble.
  4. [Section 2.3, semantic preservation metric] The CER metric used to assert that CosyVoice preserves semantic content is circular for TTS methods: CosyVoice synthesises audio from the Paraformer ASR transcript of the original speech, and the same Paraformer model is then used to transcribe the anonymised speech for CER computation. This inflates the semantic-preservation score of TTS methods by construction and weakens the complementarity claim (CosyVoice for semantics + RVC for acoustics) in Section 4.2. Please provide an external semantic evaluation, e.g., human transcription or a different ASR model, or re-frame the conclusion to acknowledge that the semantic benefit is measured through the same ASR pipeline that generated the TTS input.
minor comments (6)
  1. [Abstract and Section 5] There are typos: 'yeilds' in the Abstract and 'hightlight' in the Conclusion should be 'yields' and 'highlight'.
  2. [Fig. 1 and Section 1] The figure caption and text use inconsistent spelling: 'Anoymised' in Fig. 1 should be 'Anonymised'. Also 'ppe-bound' in Section 4.2 should be 'upper-bound'.
  3. [Table 2] The column header 'L1 PCC↑' is unclear and inconsistent with Table 1, which lists 'L1 F0↓' and 'PCC F0↑' as separate columns. Please fix the header.
  4. [Section 3.2] Please state explicitly the number of test samples (approximately 122) and clarify that all reported accuracies are on this test set. This context is essential for interpreting the differences in Fig. 4.
  5. [Fig. 4] The figure shows only point estimates with no error bars; adding error bars or shaded confidence intervals would substantially improve interpretability.
  6. [Section 2.2.2] The RVC description mentions 'retrieves segments or acoustic embeddings from a target-speaker database' but no details are given about the target database, speaker coverage, or whether the database overlaps with the evaluation speakers. Please clarify.

Circularity Check

1 steps flagged · score 2.0 of 10

Semantic-preservation CER for the TTS cascades is partly self-fulfilling because the same ASR generates the transcript and measures the error; the suicide-risk detection result itself is externally evaluated.

  1. self definitional [Section 2.2.3 (Speech Synthesis from Transcribed Text) and Section 2.3 (Preservation of Semantic Content), Table 1]
    "we investigate a cascaded approach that first conducts automatic speech recognition (ASR) to transcribe the original speech to text, and then synthesises new speech from the transcribed text using text-to-speech (TTS) technology. ... We employ Paraformer [19, 20] for ASR, and evaluate two popular TTS models: SparkTTS [21] and CosyVoice 2.0 [22]. ... The semantic consistency is evaluated by character error rate (CER) between the original and anonymised speech, both transcribed by the Paraformer ASR system."

    For the TTS cascades, the anonymised speech is synthesised from Paraformer's own transcription of the original speech. CER is then computed by transcribing both original and anonymised speech with the same Paraformer system. Because the TTS output is constructed to be a spoken rendering of that transcript, Paraformer tends to recover the same token sequence, so low CER for SparkTTS and CosyVoice is partly guaranteed by the evaluation design rather than by semantic preservation from the original signal. This inflates the Table 1 CER comparison and the statement that CosyVoice retains most semantic content, which is used to motivate the CosyVoice+RVC ensemble.

full rationale

The paper's main prediction, 0.692 ensemble accuracy versus 0.702 on original speech, is externally evaluated: the speech-LLM detector is fine-tuned on original recordings and then applied to anonymised recordings, and the anonymisation systems are pretrained off-the-shelf models with no parameters fitted to the suicide-risk labels. No fitted input is renamed as a prediction, and no load-bearing claim rests on a self-citation or an author-imported uniqueness theorem. The one genuinely circular element is the semantic-preservation CER metric for the TTS cascades: since Paraformer writes the transcript that SparkTTS/CosyVoice are asked to speak and Paraformer also computes the CER, low CER for these methods is substantially self-definitional. This weakens the complementarity story (CosyVoice preserves semantics, RVC preserves prosody) but does not reduce the central accuracy result to its inputs. The under-specified EER protocol and the high EER of original speech (0.185) are serious measurement concerns for the privacy claim, but they are not a circularity of the argument's derivation chain.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper's central trade-off claim rests on several domain assumptions: MINI-KID labels are valid, EER reflects true identity protection, the speech LLM captures risk-relevant cues, and the emotion and ASR metrics proxy semantic and affective preservation. The manual configuration selections are the main hand-chosen parameters. No new entities are introduced.

free parameters (2)
  • DoRA rank and alpha = rank=32, alpha=64
    Chosen by hand for parameter-efficient fine-tuning of Qwen2.5-Omni-7B (Section 2.1). Affects detection accuracy on all conditions but is not central to the anonymisation comparison.
  • Anonymisation configuration per method = e.g., pitch step=4, McAdams LPC=20/25, SSL-SAS with F0, FreeVC child, SeedVC d25, RVC s1
    Selected from small grids based on the quality-versus-anonymisation trade-off in Table 1 (Section 4.1). The hand-selection is based on observed metrics, so it is not independent of the evaluation framework.
assumptions (5)
  • domain assumption MINI-KID suicidality module provides valid ground-truth suicide-risk labels for the adolescent cohort.
    Section 3.1 treats MINI-KID assessment as the reference standard without external validation in this study.
  • domain assumption EER as used here measures the risk that an anonymised utterance can be linked to the original speaker.
    Section 2.3 defines EER as the speaker-discrimination metric but does not specify the verification model or protocol; the high EER (0.185) on original speech is unexplained.
  • domain assumption The Qwen2.5-Omni-7B speech LLM, fine-tuned with DoRA, captures the acoustic and semantic cues that support suicide-risk detection.
    Section 2.1 assumes this backbone is a suitable detector; no comparison with other backbones is provided in this paper.
  • domain assumption Emotion2Vec cosine similarity is a valid proxy for preservation of suicide-risk-relevant emotional content.
    Section 2.3 uses Emotion2Vec embeddings to quantify emotional preservation; no validation against human judgement or task-specific labels is reported.
  • domain assumption Paraformer ASR transcription is an adequate representation of semantic content for both synthesis input and evaluation.
    Sections 2.2.3 and 2.3 use the same ASR to create TTS text and to compute CER, so semantic-preservation scores for TTS methods are partly self-consistent.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Speaker Anonymisation for Speech-based Suicide Risk Detection." pith.science (2026). https://pith.science/paper/OIIQ57KB

@misc{pith2026250922148,
  author       = {Pith},
  title        = {Pith review of: Speaker Anonymisation for Speech-based Suicide Risk Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OIIQ57KB}},
  note         = {Machine review of arXiv:2509.22148}
}
read the original abstract

Adolescent suicide is a critical global health issue, and speech provides a cost-effective modality for automatic suicide risk detection. Given the vulnerable population, protecting speaker identity is particularly important, as speech itself can reveal personally identifiable information if the data is leaked or maliciously exploited. This work presents the first systematic study of speaker anonymisation for speech-based suicide risk detection. A broad range of anonymisation methods are investigated, including techniques based on traditional signal processing, neural voice conversion, and speech synthesis. A comprehensive evaluation framework is built to assess the trade-off between protecting speaker identity and preserving information essential for suicide risk detection. Results show that combining anonymisation methods that retain complementary information yields detection performance comparable to that of original speech, while achieving protection of speaker identity for vulnerable populations.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 27 canonical work pages

  1. [1]

    Speaker Anonymisation for Speech-based Suicide Risk Detection

    INTRODUCTION Suicide is a critical global health challenge and is one of the leading causes of death among adolescents [1, 2]. Early detection of suicide risk is essential for effective prevention and intervention of potential suicide attempts. Existing approaches still face limitations: clinical interviews require well-trained professionals, while self-r...

    work page Pith review arXiv 2026

  2. [2]

    Speech-based Suicide Risk Detection This section describes the structure used in this study for speech- based suicide risk detection

    METHODS 2.1. Speech-based Suicide Risk Detection This section describes the structure used in this study for speech- based suicide risk detection. The task is formulated as binary classification. As shown in Fig. 2, a speech LLM (Qwen2.5-Omni- 7B [14]1), which consists of a speech encoder and a large language model (LLM), is employed as the backbone for p...

    work page

  3. [3]

    Dataset The dataset used in this study consists of voice recordings collected from 1,223 Chinese adolescents aged 10-18 [8, 4]

    EXPERIMENTAL SETUP 3.1. Dataset The dataset used in this study consists of voice recordings collected from 1,223 Chinese adolescents aged 10-18 [8, 4]. All recordings were conducted in Mandarin Chinese. Recordings are all paired with standardised suicide risk assessments administered via the Mini In- ternational Neuropsychiatric Interview for Children and...

    work page

  4. [4]

    RESULTS AND DISCUSSION 4.1. Anonymisation Effectiveness and Perceptual Quality Table 1 compares various anonymisation methods across different configurations under the evaluation framework introduced in Sec- tion 2.3. We also provide a visualisation of changes in F0 contours induced by different anonymisation methods in Fig. 3. Firstly, methods that direc...

    work page arXiv 2020

  5. [5]

    CONCLUSION This work presents the first systematic investigation of speaker anonymisation methods for adolescent suicide risk detection, cov- ering traditional signal processing, neural voice conversion, and speech synthesis approaches under a comprehensive evaluation framework. Results show that different anonymisation approaches excel at preserving diff...

    work page

  6. [6]

    World Health Organization,Suicide worldwide in 2021: Global health estimates, World Health Organization, 2025

    work page 2021

  7. [7]

    Suicide and suicide attempts in adolescents,

    Benjamin Shain, Committee on Adolescence, Paula K Braver- man, William P Adelman, Elizabeth M Alderman, Cora C Bre- uner, David A Levine, Arik V Marcell, and Rebecca F O’Brien, “Suicide and suicide attempts in adolescents,”Pediatrics, vol. 138, no. 1, pp. e20161420, 2016

    work page 2016

  8. [8]

    Trust is the basis for effective suicide risk screening and as- sessment in veterans,

    Linda Ganzini, Lauren M Denneson, Nancy Press, Matthew J Bair, Drew A Helmer, Jennifer Poat, and Steven K Dobscha, “Trust is the basis for effective suicide risk screening and as- sessment in veterans,”Journal of general internal medicine, vol. 28, no. 9, pp. 1215–1221, 2013

    work page 2013

Show all 31 references
  1. [9]

    The 1st SpeechWellness challenge: Detecting suicide risk among adolescents,

    Wen Wu, Ziyun Cui, Chang Lei, Yinan Duan, Diyang Qu, Ji Wu, Bowen Zhou, Runsen Chen, and Chao Zhang, “The 1st SpeechWellness challenge: Detecting suicide risk among adolescents,” inProc. Interspeech, Rotterdam, 2025

    2025

  2. [10]

    A review of depression and suicide risk assessment using speech analysis,

    Nicholas Cummins, Stefan Scherer, Jarek Krajewski, Sebastian Schnieder, Julien Epps, and Thomas F Quatieri, “A review of depression and suicide risk assessment using speech analysis,” Speech communication, vol. 71, pp. 10–49, 2015

    2015

  3. [11]

    Read speech voice quality and disfluency in individuals with recent suicidal ideation or suicide attempt,

    Brian Stasak, Julien Epps, Heather T Schatten, Ivan W Miller, Emily Mower Provost, and Michael F Armey, “Read speech voice quality and disfluency in individuals with recent suicidal ideation or suicide attempt,”Speech Communication, vol. 132, pp. 10–20, 2021

    2021

  4. [12]

    Acoustic and language analysis of speech for suicidal ideation among us veterans,

    Anas Belouali, Samir Gupta, Vaibhav Sourirajan, Jiawei Yu, Nathaniel Allen, Adil Alaoui, Mary Ann Dutton, and Matthew J Reinhard, “Acoustic and language analysis of speech for suicidal ideation among us veterans,”BioData min- ing, vol. 14, no. 1, pp. 11, 2021

    2021

  5. [13]

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

    Ziyun Cui, Chang Lei, Wen Wu, Yinan Duan, Diyang Qu, Ji Wu, Runsen Chen, and Chao Zhang, “Spontaneous speech- based suicide risk detection using Whisper and large language models,” inProc. Interspeech, Kos Island, 2024

    2024

  6. [14]

    Anonymizing and sharing med- ical text records,

    Xiao-Bai Li and Jialun Qin, “Anonymizing and sharing med- ical text records,”Information Systems Research, vol. 28, no. 2, pp. 332–352, 2017

    2017

  7. [15]

    GANonymization: A GAN-based face anonymization framework for preserving emotional expres- sions,

    Fabio Hellmann, Silvan Mertes, Mohamed Benouis, Alexander Hustinx, Tzung-Chien Hsieh, Cristina Conati, Peter Krawitz, and Elisabeth Andr ´e, “GANonymization: A GAN-based face anonymization framework for preserving emotional expres- sions,”ACM Transactions on Multimedia Computi...

    2024

  8. [16]

    PrivECG: Generating private ECG for end-to-end anonymiza- tion,

    Alexis Nolin-Lapalme, Robert Avram, and Hussin Julie, “PrivECG: Generating private ECG for end-to-end anonymiza- tion,” inProc. MLHC, New York, 2023

    2023

  9. [17]

    Your substance abuse disorder is an open secret! Gleaning sensitive personal infor- mation from templates in an EEG-based authentication sys- tem,

    Richard Matovu and Abdul Serwadda, “Your substance abuse disorder is an open secret! Gleaning sensitive personal infor- mation from templates in an EEG-based authentication sys- tem,” inProc. BTAS, Niagara Falls, 2016

    2016

  10. [18]

    Addressing challenges in speaker anonymization to maintain utility while ensuring privacy of pathological speech,

    Soroosh Tayebi Arasteh, Tom ´as Arias-Vergara, Paula Andrea P´erez-Toro, Tobias Weise, Kai Packh¨auser, Maria Schuster, El- mar Noeth, Andreas Maier, and Seung Hee Yang, “Addressing challenges in speaker anonymization to maintain utility while ensuring privacy of pathological ...

    2024

  11. [19]

    Qwen2.5-Omni technical report,

    Jin Xu, Zhifang Guo, Jinzheng He, Hangrui Hu, Ting He, Shuai Bai, Keqin Chen, Jialin Wang, Yang Fan, Kai Dang, et al., “Qwen2.5-Omni technical report,”arXiv preprint arXiv:2503.20215, 2025

    2025 arXiv

  12. [20]

    Speaker anonymisation using the McAdams coefficient,

    Jose Patino, Natalia Tomashenko, Massimiliano Todisco, An- dreas Nautsch, and Nicholas Evans, “Speaker anonymisation using the McAdams coefficient,” inProc. Interspeech, Brno, 2021

    2021

  13. [21]

    Analyzing language-independent speaker anonymization framework under unseen conditions,

    Xiaoxiao Miao, Xin Wang, Erica Cooper, Junichi Yamagishi, and Natalia Tomashenko, “Analyzing language-independent speaker anonymization framework under unseen conditions,” inProc. Interspeech, Incheon, 2022

    2022

  14. [22]

    FreeVC: Towards high- quality text-free one-shot voice conversion,

    Jingyi Li, Weiping Tu, and Li Xiao, “FreeVC: Towards high- quality text-free one-shot voice conversion,” inProc. ICASSP, Rhodes, 2023

    2023

  15. [23]

    Zero-shot voice conversion with diffusion transformers,

    Songting Liu, “Zero-shot voice conversion with diffusion transformers,”arXiv preprint arXiv:2411.09943, 2024

    2024 arXiv

  16. [24]

    Paraformer: Fast and accurate parallel Transformer for non- autoregressive end-to-end speech recognition,

    Zhifu Gao, ShiLiang Zhang, Ian McLoughlin, and Zhijie Yan, “Paraformer: Fast and accurate parallel Transformer for non- autoregressive end-to-end speech recognition,” inProc. Inter- speech, Incheon, 2022

    2022

  17. [25]

    FunASR: A fundamental end-to-end speech recognition toolkit,

    Zhifu Gao, Zerui Li, Jiaming Wang, Haoneng Luo, Xian Shi, Mengzhe Chen, Yabin Li, Lingyun Zuo, Zhihao Du, and Shil- iang Zhang, “FunASR: A fundamental end-to-end speech recognition toolkit,” inProc. Interspeech, Dublin, 2023

    2023

  18. [26]

    Spark-TTS: An efficient LLM-based text-to-speech model with single-stream decou- pled speech tokens,

    Xinsheng Wang, Mingqi Jiang, Ziyang Ma, Ziyu Zhang, Songxiang Liu, Linqin Li, Zheng Liang, Qixi Zheng, Rui Wang, Xiaoqin Feng, et al., “Spark-TTS: An efficient LLM-based text-to-speech model with single-stream decou- pled speech tokens,”arXiv preprint arXiv:2503.01710, 2025

    2025 arXiv

  19. [27]

    CosyV oice 2: Scalable streaming speech synthesis with large language models,

    Zhihao Du, Yuxuan Wang, Qian Chen, Xian Shi, Xiang Lv, Tianyu Zhao, Zhifu Gao, Yexin Yang, Changfeng Gao, Hui Wang, et al., “CosyV oice 2: Scalable streaming speech synthesis with large language models,”arXiv preprint arXiv:2412.10117, 2024

    2024 arXiv

  20. [28]

    UTMOS: UTokyo-SaruLab system for V oiceMOS Challenge 2022,

    Takaaki Saeki, Detai Xin, Wataru Nakata, Tomoki Koriyama, Shinnosuke Takamichi, and Hiroshi Saruwatari, “UTMOS: UTokyo-SaruLab system for V oiceMOS Challenge 2022,” in Proc. Interspeech, Incheon, 2022

    2022

  21. [29]

    Emotion2Vec: Self- supervised pre-training for speech emotion representation,

    Ziyang Ma, Zhisheng Zheng, Jiaxin Ye, Jinchao Li, Zhifu Gao, ShiLiang Zhang, and Xie Chen, “Emotion2Vec: Self- supervised pre-training for speech emotion representation,” in Proc. ACL, Bangkok, 2024

    2024

  22. [30]

    Reliability and va- lidity of the mini international neuropsychiatric interview for children and adolescents (MINI-KID),

    David V Sheehan, Kathy H Sheehan, R Douglas Shytle, Juris Janavs, Yvonne Bannon, Jamison E Rogers, Karen M Milo, Saundra L Stock, and Berney Wilkinson, “Reliability and va- lidity of the mini international neuropsychiatric interview for children and adolescents (MINI-KID),”The...

    2010

  23. [31]

    FRCRN: Boosting feature representation using frequency recurrence for monaural speech enhancement,

    Shengkui Zhao, Bin Ma, Karn N Watcharasupat, and Woon- Seng Gan, “FRCRN: Boosting feature representation using frequency recurrence for monaural speech enhancement,” in Proc. ICASSP, Singapore, 2022. 5

    2022

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.