Pith. sign in

REVIEW 3 major objections 4 minor 34 references

Pitch-and-Spectrum-Aware Singing Quality Assessment with Bias Correction and Model Fusion

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Pitch and spectrum features top singing-quality benchmark

desk verdict Strong held-out benchmark result for singing MOS prediction, but the 'significantly outperforms' claim needs uncertainty quantification before it can be taken at face value. read the letter →

arxiv 2411.11123 v3 pith:SAYUBYHP submitted 2024-11-17 cs.SD eess.AS

Yu-Fei Shi , Yang Ai , Ye-Xin Lu , Hui-Peng Du , Zhen-Hua Ling This is my paper · ORCID
classification cs.SDeess.AS
keywords singingqualityassessmentMOSpredictionpitchhistogramself-supervisedlearningbiascorrectionmodelfusionAPCodecVoiceChallenge
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes PS-SQA, an automatic singing quality assessment method that predicts the mean opinion score (MOS) of synthesized singing. It argues that existing SSL-based MOS predictors, built for speech, overlook the two signals that matter most for singing: pitch (melody) and spectral detail. PS-SQA therefore conditions SSL predictors on a pitch histogram or compressed pitch and on spectral features from a non-quantized neural codec, adds a bias-correction branch that separately adjusts high- and low-score predictions to counter imbalanced training labels, and fuses the top five predictors. On the official evaluation set of the VoiceMOS 2024 Track 2 benchmark, PS-SQA reports a system-level SRCC of 0.888, surpassing the official baseline's 0.859 and every participating system. If the result holds, automatic quality assessment of singing voice synthesis and conversion becomes practical without subjective listening tests.

What carries the argument

The central machinery has four parts. First, pitch-aware conditioning: detected pitch is converted from Hz to cents and folded modulo one octave into 120 bins; the resulting per-frame compressed pitch is concatenated with SSL frame features, or the utterance-level pitch histogram is concatenated after mean pooling. Second, spectrum-aware conditioning: a non-quantized APCodec encoder, which codes amplitude and phase spectra, extracts 64-dimensional spectral features that pass through a two-layer Conformer and are concatenated with SSL features. Third, a bias-correction branch: two linear branches learn additive corrections for predicted scores above a threshold $\alpha$ and below a threshold $\beta$, leaving the middle range untouched, and are trained while the rest of the model is frozen. Fourth, model fusion: outputs of the top five predictors, ranked by system-level SRCC on the validation set, are concatenated and mapped through a linear layer to the final MOS score.

What would settle it

Retrain the full PS-SQA pipeline on a different singing MOS dataset, or on SingMOS with the validation set excluded from all threshold and top-five selection, and recompute system-level SRCC; the central claim fails if the fused, bias-corrected model no longer clearly beats the best individual predictor and the official baseline. A cheaper check is to vary $\alpha$ and $\beta$ across their allowed range, $1<\beta<\alpha<5$, and test whether system-level SRCC on the evaluation set is stable or peaks only at the validation-selected values.

Watch

Extended reading notes

Core claim

The paper's central claim is that explicitly injecting singing-specific pitch and spectrum information into an SSL-based MOS predictor, along with a bias-correction branch and model fusion, yields the state-of-the-art automatic predictor of synthesized singing quality. On the SingMOS evaluation set, the proposed PS-SQA achieves a system-level SRCC of 0.888, beating the official baseline (0.859) and all participating teams across system-level metrics. The authors further show that pitch-histogram conditioning is the most reliable single ingredient: three of the five fused predictors use it, and it outperforms plain SSL predictors on most metrics across five different SSL backbones, whereas compressed-pitch conditioning is sensitive to the backbone. The bias-correction branch reduces prediction error in low-MOS segments where training samples are scarce, and fusion of the top five predictors adds another 0.013 of system-level SRCC over the best individual predictor.

Load-bearing premise

The whole gain rests on choices made on the 544-sample validation set, namely the bias-correction thresholds $\alpha$ and $\beta$ from Eq. (6), whose values are never reported, and the selection of the top five predictors; with no sensitivity analysis, validation-specific overfit could make the reported test-set improvement shrink or invert.

Editorial extensions

If this is right

  • Automatic MOS prediction for synthesized singing can be built on SSL predictors by adding pitch and spectrum conditioning; pitch histograms are the more reliable injection method across SSL backbones.
  • The bias-correction branch can be attached to any MOS predictor's output, offering a general fix for training label imbalance without retraining the base model.
  • Fusing a handful of top-ranked predictors, ranked by validation SRCC, improves accuracy over any single predictor, so the method can be improved by adding more predictor types.
  • System-level SRCC becomes the practical optimization target: PS-SQA's gains concentrate at the system level, with 0.888 versus the baseline's 0.859.
  • The approach enables rapid, reference-free evaluation of singing voice synthesis and conversion systems, replacing time-consuming subjective listening tests.

Reading between the lines

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

  • The same recipe, validation-set model selection plus per-segment bias correction, may transfer to other low-resource subjective rating tasks such as speech MOS or music generation evaluation, where label distributions are similarly skewed toward mid-range scores.
  • The pitch-histogram success suggests that other octave-folding music features, such as chroma or key-invariant representations, could further improve singing assessment by capturing harmony rather than absolute pitch.
  • Because the paper never reports the values of $\alpha$ and $\beta$, a natural follow-up is an ablation that sweeps these thresholds; if the optimal thresholds vary across datasets, the bias-correction branch needs a data-driven or adaptive estimation step.
  • The non-quantized APCodec feature stream is a distinct contribution: discarding quantization for quality assessment rather than compression means the same encoder could serve other audio-quality tasks that need fine spectral detail.
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

3 major / 4 minor

Summary. The paper proposes PS-SQA, a singing mean-opinion-score (MOS) prediction system for VoiceMOS 2024 Track 2. It extends plain SSL-based MOS predictors by conditioning on pitch histograms or compressed pitch sequences and by appending non-quantized APCodec spectral features, adds a bias-correction branch with two linear correction branches selected by thresholds, and fuses the top five predictors chosen by validation-set system-level SRCC. The authors report that their original submission placed first among non-baseline participants and that the improved PS-SQA achieves system-level SRCC 0.888 on the official evaluation set, compared with 0.859 for the official baseline and 0.856 for their submitted system.

Significance. The paper's empirical contribution is useful for the singing-quality-assessment community: it provides a systematic comparison of 20 SSL-based predictor variants on a public challenge dataset, demonstrates that pitch-histogram conditioning is more effective than compressed-pitch conditioning, and introduces a simple and plausible bias-correction mechanism for low-resource MOS intervals. The reported point estimates suggest that PS-SQA is competitive with, and likely superior to, the official baseline and other participants on this benchmark. However, the central claim that PS-SQA 'significantly outperforms' all participating systems is not backed by confidence intervals, significance tests, or even a statement of the number of systems over which system-level SRCC is computed. In addition, the key bias-correction thresholds are never reported, and the fusion gain is evaluated on the same validation set used to select the fused predictors, which introduces optimistic bias. These issues are fixable but currently limit the strength of the conclusions.

major comments (3)
  1. [§5, Table 3] The claim that PS-SQA 'significantly outperforms all participating systems in terms of system-level evaluation metrics' is supported only by point estimates in Table 3. The number of systems used to compute system-level SRCC is not stated; Table 3 lists eight rows (seven baselines/participants plus PS-SQA), so the correlation is computed over a very small sample. With roughly 7–8 systems, a difference of 0.029 (0.888 vs. 0.859) may fall well within sampling variability. Please provide confidence intervals from bootstrap resampling over systems, a significance test, or explicitly soften the wording to 'outperforms in terms of point estimates.' Without this, the abstract and Section 5 overstate the evidence.
  2. [§3.3, Eq. (6)] The thresholds α and β in the bias-correction branch are never reported, nor is the criterion for choosing them described. Since the bias-correction branch is a core contribution and the final PS-SQA system in Table 3 uses it, the method is not reproducible without these values. Please report α and β, state how they were selected (e.g., grid search on the validation set), and describe the training procedure for the addition and subtraction branches after the base model is frozen.
  3. [§4.3, Table 2] The top five predictors are selected by ranking their system-level SRCC on the same 544-sample validation set that is then used to evaluate the fusion and bias-correction gains in Table 2. This selection-on-validation-set makes the reported improvements (e.g., fusion improving system-level SRCC by 0.013 over the best individual predictor) optimistically biased and not independent evidence. The evaluation-set results in Table 3 are less affected, but the paper should either use a nested or separate selection procedure to demonstrate the fusion gain, or explicitly acknowledge that the fusion improvement is measured with a selection bias on the validation set.
minor comments (4)
  1. [§4.2] The heading 'Comparsion among Different SSL-based MOS Predictors' contains a typo; it should be 'Comparison.'
  2. [§4.2] The checkpoint-selection sentence, 'If the system-level SRCC didn’t decrease within 15 epochs, early stopping was applied,' is ambiguous because a higher SRCC is better. Please clarify whether early stopping is triggered when SRCC fails to improve (increase) for 15 epochs.
  3. [Table 3 caption] The table caption does not state the number of systems over which system-level metrics are computed; please add this information to the caption or the text, as it is essential for interpreting the system-level SRCC values.
  4. [§4.4] The statement that the submitted system (T08) 'significantly outperformed' the other participating systems has the same statistical-support issue as the main claim; please either add uncertainty quantification or use more cautious phrasing throughout.

Circularity Check

1 steps flagged · score 3.0 of 10

Mild validation-set selection bias: the top-five fusion predictors are selected by system-level SRCC on the validation set, and the reported fusion gain is measured on that same validation set; the held-out test-set claim remains independent.

  1. fitted input called prediction [Section 4.3, Table 2]
    "Since system-level SRCC is an important ranking metric in V oiceMOS Challenge 2024, we selected the top five predictors based on their system-level SRCC rankings. ... Specifically, after model fusion, the system-level SRCC improved by 0.013 compared to the best individual predictor. This indicates that the model fusion strategy combines the strengths of individual MOS predictors, achieving more accurate singing MOS predictions. This confirms the effectiveness of model fusion."

    The selection criterion for the five predictors is system-level SRCC computed on the SingMOS validation set, and the evidence for the fusion gain is the system-level SRCC of the fused model computed on the same validation set. The reported 0.013 improvement over the best individual predictor is therefore not an independent estimate: choosing the top five predictors by the same metric used to evaluate the ensemble inflates the apparent gain by construction. The bias-correction thresholds alpha and beta in Eq. (6) are also part of the validation-set configuration, and Table 2 reports with/without bias-correction comparisons on the same validation set, adding a similar selection bias.

full rationale

The paper's load-bearing claim—that PS-SQA significantly outperforms all participating systems on VoiceMOS 2024 Track 2—rests on Table 3, which reports system-level SRCC on the held-out evaluation set. Although architecture choices (pitch-histogram conditioning, non-quantized APCodec features, bias-correction branch, fusion) were selected using the 544-sample validation set, the evaluation-set numbers are not used to fit anything, so the central claim has independent empirical content. No theoretical derivation is bootstrapped: Eq. (6) defines the bias-corrected score from the original prediction plus branch outputs, and the pitch-histogram and APCodec features are external inputs rather than outputs of the MOS model. The only circularity-adjacent issue is in Section 4.3: the top-five predictors are selected by validation system-level SRCC, and the same validation system-level SRCC is then reported as evidence that fusion improves SRCC by 0.013; this is a selection-on-evaluation-set artifact, not an independent confirmation. It does not invalidate the held-out test comparison, but it means the validation-set fusion and bias-correction gains in Table 2 are optimistically biased. The APCodec citation is same-group prior work, but it is used as a fixed pretrained codec, so the self-citation is not load-bearing. Score 3 reflects one mild fitted-input-called-prediction step on the validation set; the central competition result is not circular.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central system depends on a large set of pretrained components (SSL models, APCodec) taken from prior work, and on validation-based choices (alpha, beta, predictor selection) that are not fully specified. There are no new physical entities; the contribution is an architecture and training scheme.

free parameters (3)
  • Bias correction thresholds alpha and beta = unreported
    Used in Eq. (6) to decide which branch corrects a predicted MOS; values are not given and appear hand-tuned to the SingMOS validation set, directly affecting final predicted scores and the reported bias-correction improvements.
  • Fusion predictor selection (top 5 of 20) = CP-Wav2Vec2.0 Base, PH-Wav2Vec2.0 Base, PH-Wav2Vec2.0 Large, PH-HuBERT Base, S-HuBERT Base
    The number and identity of predictors to fuse are chosen after inspecting validation-set system-level SRCC rankings (Section 4.3), so the fusion architecture is fitted to the validation distribution.
  • Bias correction training setup = not specified
    The paper says only the two bias branches are trained after the main model is fixed, but gives no learning rate, number of epochs, or early stopping for this stage; these choices affect the final predictions and are not reported.
assumptions (5)
  • domain assumption Ground-truth MOS labels in SingMOS are valid, consistent, and reflect singing quality.
    All training and evaluation rely on these labels as the target; introduced in Section 4.1.
  • domain assumption SSL features pretrained on speech transfer to singing voice quality assessment.
    The plain SSL-based MOS predictor (Section 2.1) fine-tunes speech SSL models on singing MOS labels; no evidence that SSL representations capture singing-relevant acoustics is provided beyond the empirical results.
  • domain assumption Pitch histograms, folded into one octave, capture the melodic accuracy that determines singing quality.
    Motivated in Section 2.2 with reference to prior work [1,14] and Figure 2; the pitch-aware predictors depend on this.
  • domain assumption Non-quantized APCodec encoder features, despite being trained on the VCTK English speech corpus, provide spectral cues useful for Mandarin and Japanese singing MOS prediction.
    Section 3.2 uses a non-quantized APCodec pretrained on VCTK-0.92 [26] as a fixed feature extractor; no singing-domain fine-tuning or checks of domain shift are reported.
  • domain assumption The validation set is representative enough for selecting checkpoints, predictors, and bias-correction thresholds.
    All model selection decisions (Sections 4.2 and 4.3) are made on the 544-sample validation set; the test set is used once.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Pitch-and-Spectrum-Aware Singing Quality Assessment with Bias Correction and Model Fusion." pith.science (2026). https://pith.science/paper/SAYUBYHP

@misc{pith2026241111123,
  author       = {Pith},
  title        = {Pith review of: Pitch-and-Spectrum-Aware Singing Quality Assessment with Bias Correction and Model Fusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SAYUBYHP}},
  note         = {Machine review of arXiv:2411.11123}
}
read the original abstract

We participated in track 2 of the VoiceMOS Challenge 2024, which aimed to predict the mean opinion score (MOS) of singing samples. Our submission secured the first place among all participating teams, excluding the official baseline. In this paper, we further improve our submission and propose a novel Pitch-and-Spectrum-aware Singing Quality Assessment (PS-SQA) method. The PS-SQA is designed based on the self-supervised-learning (SSL) MOS predictor, incorporating singing pitch and spectral information, which are extracted using pitch histogram and non-quantized neural codec, respectively. Additionally, the PS-SQA introduces a bias correction strategy to address prediction biases caused by low-resource training samples, and employs model fusion technology to further enhance prediction accuracy. Experimental results confirm that our proposed PS-SQA significantly outperforms all competing systems across all system-level metrics, confirming its strong sing quality assessment capabilities.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 31 canonical work pages

  1. [1]

    Pitch-and-Spectrum-Aware Singing Quality Assessment with Bias Correction and Model Fusion

    INTRODUCTION With the rapid development of singing voice synthesis (SVS) and singing voice conversion (SVC) systems, there is an urgent need for technology that can automatically assess the quality of generated singing voice, instead of traditional subjective listener scoring meth- ods which are time-consuming and inefficient. However, in past re- search ...

    work page Pith review arXiv 2024

  2. [2]

    RELA TED WORK 2.1. SSL-based MOS Predictor Recently, SSL models trained with a large amount of unlabeled data using self-supervised learning have been applied to MOS prediction, achieving impressive results [9]. In both the V oiceMOS Challenge 2022 and 2023, top-ranking teams employed fine-tuning on SSL models to achieve perfect predictive accuracy [10, 1...

    work page 2022

  3. [3]

    Finally, a linear layer reduces the feature dimensionality to 1 to derive the corresponding MOS score

    The waveform is processed through a pre-trained SSL model to produce a frame-level feature vector, which is then averaged using mean-pooling to obtain an utterance-level one. Finally, a linear layer reduces the feature dimensionality to 1 to derive the corresponding MOS score. Assuming ˆy is the predicted MOS and y is its corre- sponding label, the loss f...

    work page 2024

  4. [4]

    Additionally, to overcome the issues caused by imbalanced training data, PS-SQA innovatively introduces a bias correction strategy

    PROPOSED METHOD The core of PS-SQA lies in introducing pitch-aware SSL-based MOS predictors and spectrum-aware SSL-based MOS predictors based on the plain SSL-based MOS prediction framework, tailored for quality assessment that suits the characteristics of singing voices. Additionally, to overcome the issues caused by imbalanced training data, PS-SQA inno...

    work page 2024

  5. [5]

    Dataset and Evaluation Metrics During the training phase, the track 2 of the V oiceMOS Challenge 2024 released a dataset named SingMOS[25]

    EXPERIMENTS 4.1. Dataset and Evaluation Metrics During the training phase, the track 2 of the V oiceMOS Challenge 2024 released a dataset named SingMOS[25]. This dataset com- prises Mandarin and Japanese samples obtained from SVS systems, SVC systems, analysis-synthesis operation of neural vocoders, and natural singing voice recordings. In total, the data...

    work page 2024

  6. [6]

    Bias Correction

    compressed-pitch-aware MOS predictor with Wav2Vec2.0 Base as the SSL model, 2) pitch-histogram-aware MOS predictor with Wav2Vec2.0 Base as the SSL model, 3) pitch-histogram-aware MOS predictor with Wav2Vec2.0 Large as the SSL model, 4) pitch- histogram-aware MOS predictor with HuBERT Base as the SSL model, and 5) spectrum-aware MOS predictor with HuBERT B...

    work page 2024

  7. [7]

    CONCLUSION This paper proposes a novel pitch-and-spectrum-aware singing qual- ity assessment method, called PS-SQA, which is an improvement version of the system we submitted to track 2 of V oiceMOS Chal- lenge 2024. The PS-SQA first introduces multiple MOS predic- tors that incorporate pitch and spectrum-related information into the SSL-based MOS predict...

    work page 2024

  8. [8]

    Spectral fea- tures and pitch histogram for automatic singing quality evalu- ation with CRNN,

    Lin Huang, Chitralekha Gupta, and Haizhou Li, “Spectral fea- tures and pitch histogram for automatic singing quality evalu- ation with CRNN,” in Proc. APSIPA, 2020, pp. 492–499

    work page 2020

Show all 34 references
  1. [9]

    Training ex- plainable singing quality assessment network with augmented data,

    Jinhu Li, Chitralekha Gupta, and Haizhou Li, “Training ex- plainable singing quality assessment network with augmented data,” in Proc. APSIPA, 2021, pp. 904–911

    2021

  2. [10]

    Tg-Critic: A timbre-guided model for reference-independent singing evaluation,

    Xiaoheng Sun, Yuejie Gao, Hanyao Lin, and Huaping Liu, “Tg-Critic: A timbre-guided model for reference-independent singing evaluation,” in Proc. ICASSP, 2023, pp. 1–5

    2023

  3. [11]

    The Blizzard Challenge 2005: Evaluating corpus-based speech synthesis on common databases,

    A Black and Keiichi Tokuda, “The Blizzard Challenge 2005: Evaluating corpus-based speech synthesis on common databases,” in Proc. interspeech, 2005, pp. 77–80

    2005

  4. [12]

    AutoMOS: Learn- ing a non-intrusive assessor of naturalness-of-speech,

    Brian Patton, Yannis Agiomyrgiannakis, Michael Terry, Kevin Wilson, Rif A Saurous, and D Sculley, “AutoMOS: Learn- ing a non-intrusive assessor of naturalness-of-speech,” arXiv preprint arXiv:1611.09207, 2016

    2016 arXiv

  5. [13]

    Quality-Net: An end-to-end non-intrusive speech quality as- sessment model based on BLSTM,

    Szu-wei Fu, Tsao Yu, Hsin-Te Hwang, and Hsin-Min Wang, “Quality-Net: An end-to-end non-intrusive speech quality as- sessment model based on BLSTM,” in Proc. Interspeech , 2018, pp. 1873–1877

    2018

  6. [14]

    MOSNet: Deep learning-based objective assessment for voice conver- sion,

    Chen-Chou Lo, Szu-Wei Fu, Wen-Chin Huang, Xin Wang, Ju- nichi Yamagishi, Yu Tsao, and Hsin-Min Wang, “MOSNet: Deep learning-based objective assessment for voice conver- sion,” in Proc. Interspeech, 2019, pp. 1541–1545

    2019

  7. [15]

    APCodec: A neural audio codec with par- allel amplitude and phase spectrum encoding and decoding,

    Yang Ai, Xiao-Hang Jiang, Ye-Xin Lu, Hui-Peng Du, and Zhen-Hua Ling, “APCodec: A neural audio codec with par- allel amplitude and phase spectrum encoding and decoding,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 32, pp. 3256–3269, 2024

    2024

  8. [16]

    Generalization ability of MOS prediction networks,

    Erica Cooper, Wen-Chin Huang, Tomoki Toda, and Junichi Ya- magishi, “Generalization ability of MOS prediction networks,” in Proc. ICASSP, 2022, pp. 8442–8446

    2022

  9. [17]

    The V oiceMOS Chal- lenge 2022,

    Wen-Chin Huang, Erica Cooper, Yu Tsao, Hsin-Min Wang, Tomoki Toda, and Junichi Yamagishi, “The V oiceMOS Chal- lenge 2022,” in Proc. Interspeech, 2022, pp. 4536–4540

    2022

  10. [18]

    The V oiceMOS Chal- lenge 2023: zero-shot subjective speech quality prediction for multiple domains,

    Erica Cooper, Wen-Chin Huang, Yu Tsao, Hsin-Min Wang, Tomoki Toda, and Junichi Yamagishi, “The V oiceMOS Chal- lenge 2023: zero-shot subjective speech quality prediction for multiple domains,” in Proc. ASRU, 2023, pp. 1–7

    2023

  11. [19]

    Pitch histograms in audio and symbolic music information re- trieval,

    George Tzanetakis, Andrey Ermolinskyi, and Perry Cook, “Pitch histograms in audio and symbolic music information re- trieval,” Journal of New Music Research, pp. 143–152, 2003

    2003

  12. [20]

    Perceptual evaluation of singing quality,

    Chitralekha Gupta, Haizhou Li, and Ye Wang, “Perceptual evaluation of singing quality,” in Proc. APSIPA. IEEE, 2017, pp. 577–586

    2017

  13. [21]

    Automatic evaluation of singing quality without a reference,

    Chitralekha Gupta, Haizhou Li, and Ye Wang, “Automatic evaluation of singing quality without a reference,” in Proc. APSIPA, 2018, pp. 990–997

    2018

  14. [22]

    Soundstream: An end-to- end neural audio codec,

    Neil Zeghidour, Alejandro Luebs, Ahmed Omran, Jan Skoglund, and Marco Tagliasacchi, “Soundstream: An end-to- end neural audio codec,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 30, pp. 495–507, 2021

    2021

  15. [23]

    High fidelity neural audio compression,

    Alexandre D ´efossez, Jade Copet, Gabriel Synnaeve, and Yossi Adi, “High fidelity neural audio compression,” Transactions on Machine Learning Research, 2023

    2023

  16. [24]

    Deep learning-based non- intrusive multi-objective speech assessment model with cross- domain features,

    Ryandhimas E Zezario, Szu-Wei Fu, Fei Chen, Chiou-Shann Fuh, Hsin-Min Wang, and Yu Tsao, “Deep learning-based non- intrusive multi-objective speech assessment model with cross- domain features,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 31, pp. 54–70, 2022

    2022

  17. [25]

    Automatic rank-ordering of singing vocals with twin-neural network.,

    Chitralekha Gupta, Lin Huang, and Haizhou Li, “Automatic rank-ordering of singing vocals with twin-neural network.,” in Proc. ISMIR, 2020, pp. 416–423

    2020

  18. [26]

    Conformer: Convolution- augmented transformer for speech recognition,

    Anmol Gulati, James Qin, Chung-Cheng Chiu, Niki Par- mar, Yu Zhang, Jiahui Yu, Wei Han, Shibo Wang, Zheng- dong Zhang, Yonghui Wu, et al., “Conformer: Convolution- augmented transformer for speech recognition,” in Proc. Inter- speech, 2020, pp. 5036–5040

    2020

  19. [27]

    Ensemble of deep neural network models for MOS prediction,

    Marie Kune ˇsov´a, Jind ˇrich Matou ˇsek, Jan Lehe ˇcka, Jan ˇSvec, Josef Mich ´alek, Daniel Tihelka, Martin Bul ´ın, Zden ˇek Hanzl´ıˇcek, and Mark´eta ˇRez´aˇckov´a, “Ensemble of deep neural network models for MOS prediction,” in Proc. ICASSP, 2023, pp. 1–5

    2023

  20. [28]

    Fusion of self- supervised learned models for MOS prediction,

    Zhengdong Yang, Wangjin Zhou, Chenhui Chu, Sheng Li, Raj Dabre, Raphael Rubino, and Yi Zhao, “Fusion of self- supervised learned models for MOS prediction,” in Proc. In- terspeech, 2022, pp. 5443–5447

    2022

  21. [29]

    The ZevoMOS entry to V oiceMOS Challenge 2022,

    Adriana Stan, “The ZevoMOS entry to V oiceMOS Challenge 2022,” in Proc. Interspeech 2022, 2022, pp. 4516–4520

    2022

  22. [30]

    LE-SSL-MOS: Self-supervised learning MOS prediction with listener enhancement,

    Zili Qi, Xinhui Hu, Wangjin Zhou, Sheng Li, Hao Wu, Jian Lu, and Xinkang Xu, “LE-SSL-MOS: Self-supervised learning MOS prediction with listener enhancement,” in Proc. ASRU, 2023, pp. 1–6

    2023

  23. [31]

    Fairseq: A fast, extensible toolkit for sequence modeling,

    Myle Ott, Sergey Edunov, Alexei Baevski, Angela Fan, Sam Gross, Nathan Ng, David Grangier, and Michael Auli, “Fairseq: A fast, extensible toolkit for sequence modeling,” arXiv preprint arXiv:1904.01038, 2019

    1904 arXiv

  24. [32]

    SingMOS: An extensive open-source singing voice dataset for MOS pre- diction,

    Yuxun Tang, Jiatong Shi, Yuning Wu, and Qin Jin, “SingMOS: An extensive open-source singing voice dataset for MOS pre- diction,” arXiv preprint arXiv:2406.10911, 2024

    2024 arXiv

  25. [33]

    CSTR VCTK corpus: English multi-speaker corpus for CSTR voice cloning toolkit (version 0.92),

    Junichi Yamagishi, Christophe Veaux, Kirsten MacDonald, et al., “CSTR VCTK corpus: English multi-speaker corpus for CSTR voice cloning toolkit (version 0.92),” University of Ed- inburgh. The Centre for Speech Technology Research (CSTR) , 2019

    2019

  26. [34]

    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, 2022, pp. 4521–4525

    2022

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.