Pith. sign in

REVIEW 3 major objections 5 minor 48 references

MoTAS: MoE-Guided Feature Selection from TTS-Augmented Speech for Enhanced Multimodal Alzheimer's Early Screening

T0 review · 3 major / 5 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 MoTAS combines TTS-augmented speech with MoE-guided feature selection to reach 85.71% accuracy on the ADReSSo benchmark for Alzheimer's screening.

desk verdict Useful empirical recipe undercut by a likely test-set selection of the augmentation factor; the 85.71% headline should be treated as provisional. read the letter →

arxiv 2508.20513 v1 pith:P25PAFVL submitted 2025-08-28 cs.SD cs.MM

Yongqi Shao , Binxin Mei , Cong Tan , Hong Huo , Tao Fang This is my paper · ORCID
classification cs.SDcs.MM
keywords Alzheimer'sdiseasespeech-basedscreeningtext-to-speechaugmentationmixtureofexpertsmultimodalfeaturefusionADReSSoWav2Vec2ASRtranscripts
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 that speech-based screening for Alzheimer's disease can be made markedly more accurate in small clinical datasets by combining two ideas: synthesizing additional training speech with a text-to-speech (TTS) model that preserves each speaker's voice, and using a Mixture of Experts (MoE) gating mechanism to pick which acoustic and text features are informative for each sample. The resulting system, MoTAS, is reported to reach 85.71% accuracy on the ADReSSo test set, the best figure among the baselines compared, with an Alzheimer's recall of 94.29% that matters for screening programs that want to miss few cases. Ablations attribute the gain to both components: TTS alone raises accuracy to 81.72%, MoE alone to 79.71%, and the two together to 85.71%. If this holds, the framework is a practical recipe for stretching limited clinical speech data without hand-engineering new features.

What carries the argument

The load-bearing mechanism is a two-stage feature pipeline with a gating layer in the middle. Whisper transcribes each recording, and FishSpeech synthesizes a new utterance by pairing speaker identity from one same-class sample with transcript content from another, so the synthetic audio is supposed to keep the original speaker's prosody, rhythm, and articulation while changing the words. Four feature streams are extracted: Wav2Vec2 embeddings, MFCC dynamics from a BiLSTM, ResNet18 spectrogram features, and BERT [CLS] text embeddings. The MFCC, spectrogram, and text streams each pass through an independent Mixture of Experts block—a set of specialized sub-networks whose outputs are combined by a learned gate, with three experts per feature type and a softmax-gated weighted sum $x_{ ext{MoE}} = \sum_i w_i E_i(x)$, where $E_i$ is the $i$-th expert and $w_i$ its gate weight—while Wav2Vec2 is added directly at fusion. The MoE's role is to adaptively emphasize the features most relevant to each input, and the design places Wav2Vec2 outside the gating because it already supplies rich phonetic detail.

What would settle it

Measure the distribution of pause durations, filled pauses, and articulation rate in FishSpeech-synthesized AD speech versus real AD speech. If synthetic AD samples are statistically no more dysfluent than synthetic CN samples—or are closer to synthetic CN than real AD is to real CN—then the augmentation is not preserving disease-relevant cues, and training on synthetic data is the likely source of the reported gain. A complementary check is to train with the same augmentation on one dataset and evaluate accuracy on an independent held-out set of real recordings.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that the combination of TTS data augmentation and MoE-guided multimodal feature selection yields the best result among the systems compared on the ADReSSo benchmark: 85.71% accuracy, with Alzheimer's precision/recall of 80.49%/94.29% and cognitively normal precision/recall of 93.10%/77.14%, and F1-scores of 86.84% and 84.38% respectively. The paper also reports that the augmentation effect is non-monotonic: accuracy rises from 79.71% (MoE, no TTS) to 85.71% at 2× augmentation, then falls to 82.86% at 2.5× and 80.29% at 3×, which it attributes to overfitting on the synthetic distribution when real samples are diluted. A further claim is that the MoE gating mechanism makes the model robust to ASR errors by down-weighting unreliable text features, and balances the classes better than naive fusion baselines that exhibit modal dominance.

Load-bearing premise

The argument collapses if FishSpeech, when producing speech from another speaker's same-class transcript, cleans up or normalizes the pauses, hesitations, and articulation errors that distinguish Alzheimer's from normal speech, because the augmented samples would then teach the model an unrealistic acoustic distribution.

Editorial extensions

If this is right

  • A doubling of the training set via TTS is the empirically optimal augmentation level; beyond it accuracy falls, so generating more synthetic speech does not keep helping.
  • The MoE mechanism contributes independently of augmentation, raising accuracy by about 1.4 points on the original training set.
  • The high Alzheimer's recall (94.29%) means the system, if used as a screener, would refer most true cases for follow-up while accepting a higher false-positive rate among cognitively normal speakers.
  • Because the pipeline relies on frozen pretrained encoders and a small MLP head, the classification stage is cheap enough for settings with limited computational resources.

Reading between the lines

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

  • A testable implication of the ablations is that the AD-recall gain comes largely from the synthetic samples: AD recall jumps from 76.00% (MoE without TTS) to 94.29% (MoE with 2× TTS), so an evaluation restricted to real test utterances would show how much of that gain transfers.
  • Beyond the paper, the same recipe—speaker-preserving TTS augmentation plus gated multimodal selection—is a natural fit for other low-resource clinical speech tasks, such as depression or Parkinson's screening, which face the same data-scarcity and feature-redundancy problems.
  • Beyond the paper, since each synthetic sample pairs a reference voice with another speaker's same-class transcript, the method may be learning class-typical word patterns as much as acoustic pathology; synthesizing with the speaker's own transcript would isolate acoustic from lexical contributions.
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 / 5 minor

Summary. The paper proposes MoTAS, a pipeline for speech-based Alzheimer's disease (AD) screening that combines TTS data augmentation with a Mixture-of-Experts (MoE) feature-selection module. Whisper ASR transcribes the audio, FishSpeech synthesizes additional speech by recombining each speaker's voice with same-class transcripts from other speakers, and features from Wav2Vec2, MFCC-BiLSTM, ResNet18 spectrograms, and BERT are fused after per-modality MoE gating. The method is evaluated on the ADReSSo benchmark, reporting a test accuracy of 85.71%, which exceeds the baselines listed in Table 2. Ablations in Table 3 attribute part of the gain to MoE and part to a 2x TTS augmentation factor selected via validation.

Significance. If the reported result is robust, the paper makes a useful empirical contribution to low-resource speech-based AD screening by demonstrating that TTS augmentation plus adaptive feature selection can improve over strong multimodal baselines. The architecture is simple, the ablation study is internally consistent, and the problem is practically relevant. However, the central empirical claim currently rests on a single small test set (71 subjects), a hyperparameter apparently selected on that test set, and five-run averages without error bars or significance tests. The 1.2-point margin over the best baseline is within the range of noise for this sample size. With proper validation-based model selection and uncertainty quantification, the claim could become credible; without these, the significance of the result is not established.

major comments (3)
  1. [Section 4.2, Table 3, Figure 3] The augmentation factor (2x) is selected on the ADReSSo test set, which biases the reported accuracy. Section 4.2 states the optimal ratio was chosen based on validation performance, but no validation split is described; the dataset has only 166 training and 71 test participants. Table 3 and Figure 3 report test accuracy for all candidate factors (1.5x, 2x, 2.5x, 3x), which is consistent with selecting the factor on the test set. Because the spread across factors is 80.29-85.71% (about 4 test subjects), picking the best factor and then reporting its accuracy as the headline result is optimistically biased. Please describe the exact validation protocol, or, if none exists, re-run the selection using a held-out portion of the training set and report only the test accuracy of the selected model.
  2. [Section 5.1, Table 2] The claim that MoTAS 'significantly outperforms' all baselines is not statistically supported. The gap over the best baselines (TDNN-ASR-M5 and Whisper-TL-FTP, both 84.51%) is 1.20 percentage points, which is less than one participant out of 71. The results are averaged over five seeds, but no standard deviations, confidence intervals, or significance tests are reported. Please report per-seed results and perform a paired test (e.g., McNemar on subject-level predictions) against the strongest baselines, or temper the significance claim accordingly.
  3. [Section 3.1, Eq. (2)] The TTS augmentation relies on the assumption that FishSpeech preserves disease-relevant acoustic cues such as pauses, hesitations, and articulation errors when synthesizing speech with a different transcript. This assumption is asserted but not verified. If the TTS model cleans up or normalizes dysfluencies, the augmented samples train on an unrealistic acoustic distribution, and the gain over the no-augmentation baseline may reflect the distribution of synthetic data rather than genuine AD-related cues. Please provide evidence that the synthetic speech retains these cues, for example by comparing pause statistics, articulation measures, or ASR error patterns between real and synthetic samples, or by showing that a model trained on synthetic-only data captures similar AD/CN separation as one trained on real data.
minor comments (5)
  1. [Section 4.2, Table 1] Table 1 reports the dataset split for 3x augmentation, but the final model uses 2x augmentation (Section 5.2). Please clarify the relationship between the table and the final configuration, or report the 2x split as well.
  2. [Figure 3] The figure reports test accuracy for different augmentation factors, but it would be helpful to show per-seed variability (e.g., error bars) rather than only the mean, especially given the small test set.
  3. [Section 4.2] Typographical errors: 'expended' should be 'expanded', 'settingss' should be 'settings', and 'agumentation' should be 'augmentation' (in Section 5.2 and elsewhere).
  4. [Section 3.4, Eq. (16)] The loss is written with y_i and y-hat_i, but y was defined earlier as a binary outcome in Eq. (15). Please make the notation consistent, e.g., use y_i for the true label and p_i for the predicted probability.
  5. [Section 5.1] The claim of robustness to ASR errors is interesting but is not directly tested. If the authors wish to make this claim, an experiment with corrupted or perturbed transcripts would strengthen it.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline accuracy is selected from a test-set sweep of augmentation factors, so the reported 85.71% is a fitted maximum rather than a pre-specified prediction.

  1. fitted input called prediction [Section 4.2 (Implementation Details) and Section 5.2 (Ablation Study), Table 3 and Figure 3]
    "To evaluate the impact of different augmentation scales on model performance, we conducted comparative experiments using training sets expended by 1.5×, 2×, and 2.5×. The optimal augmentation ratio was selected based on validation performance. ... Figure 3: Test Accuracy on the ADReSSo Dataset [29] With Different TTS Augmentation Factors ... Based on these findings, we selected 2× TTS augmentation as the optimal configuration."

    The ADReSSo dataset is described as having only a 166-participant training set and a 71-participant test set; no validation split is described. Table 3 and Figure 3 report accuracy on the ADReSSo test set for augmentation factors 1.5x, 2x, 2.5x, and 3x, and the 2x factor is selected because it produced the highest test accuracy (85.71%). The headline result is therefore the maximum of a test-set sweep: the augmentation factor is fitted to the test set, and the resulting best-case accuracy is then presented as the model's leading performance. With roughly 71 test participants, the 1.2-point margin over the best baseline (84.51%) is within one participant, and the claim of outperforming all baselines rests entirely on this test-selected configuration.

full rationale

This is an empirical pipeline evaluated on the ADReSSo benchmark rather than a formal derivation, and most of the paper is self-contained: there are no self-citations by the current authors, no imported uniqueness theorem, and no definitional identification of inputs with outputs. The TTS augmentation assumption that FishSpeech preserves disease-relevant prosodic and articulatory cues is unverified but is an empirical assumption, not circularity. The one load-bearing circularity concern is the selection of the TTS augmentation factor. The paper states the ratio was chosen by validation performance, but it reports test accuracy for every candidate factor in Table 3 and Figure 3 and then selects 2x because it achieved the maximum test accuracy. Consequently the central claim of a leading 85.71% accuracy reduces to a test-set-tuned hyperparameter choice, making the reported comparison against baselines optimistically biased. If a genuine held-out validation split had been described and the test accuracy reported only once after fixing the factor, this would be a normal evaluation and the circularity score would be 0-2. As written, the headline result is a fitted maximum, so the score is 6.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The main result depends on a small number of explicit design choices: the TTS preservation assumption, the augmentation factor (selected from four candidates), and the expert count k=3. No new entities are introduced. The assumption that FishSpeech preserves AD-relevant acoustic features while speaking another person's transcript is the most consequential unverified premise, since it underlies the entire augmentation strategy.

free parameters (2)
  • TTS augmentation factor = 2.0 (2x)
    Chosen from {1.5, 2, 2.5, 3} based on validation accuracy; the paper reports the resulting test accuracy (Table 3, ID 4).
  • Number of MoE experts k = 3
    Set by hand in Section 4.2; no sensitivity analysis for k.
assumptions (3)
  • domain assumption FishSpeech TTS preserves the reference speaker's disease-relevant acoustic characteristics when the linguistic content is replaced by another same-class transcript.
    Used in Section 3.1, Eq. (2), to justify that synthetic speech retains AD/CN vocal cues; if false, augmentation introduces distribution shift.
  • domain assumption Whisper transcriptions are accurate enough to serve as TTS input and text features.
    ASR errors in cognitively impaired speech could corrupt TTS content and BERT text embeddings; the paper asserts robustness but does not quantify ASR error impact.
  • domain assumption ADReSSo test set labels are correct and the train/test split is used exactly as in the challenge.
    All results depend on this; the paper does not provide independent verification of dataset labels.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MoTAS: MoE-Guided Feature Selection from TTS-Augmented Speech for Enhanced Multimodal Alzheimer's Early Screening." pith.science (2026). https://pith.science/paper/P25PAFVL

@misc{pith2026250820513,
  author       = {Pith},
  title        = {Pith review of: MoTAS: MoE-Guided Feature Selection from TTS-Augmented Speech for Enhanced Multimodal Alzheimer's Early Screening},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P25PAFVL}},
  note         = {Machine review of arXiv:2508.20513}
}
read the original abstract

Early screening for Alzheimer's Disease (AD) through speech presents a promising non-invasive approach. However, challenges such as limited data and the lack of fine-grained, adaptive feature selection often hinder performance. To address these issues, we propose MoTAS, a robust framework designed to enhance AD screening efficiency. MoTAS leverages Text-to-Speech (TTS) augmentation to increase data volume and employs a Mixture of Experts (MoE) mechanism to improve multimodal feature selection, jointly enhancing model generalization. The process begins with automatic speech recognition (ASR) to obtain accurate transcriptions. TTS is then used to synthesize speech that enriches the dataset. After extracting acoustic and text embeddings, the MoE mechanism dynamically selects the most informative features, optimizing feature fusion for improved classification. Evaluated on the ADReSSo dataset, MoTAS achieves a leading accuracy of 85.71\%, outperforming existing baselines. Ablation studies further validate the individual contributions of TTS augmentation and MoE in boosting classification performance. These findings highlight the practical value of MoTAS in real-world AD screening scenarios, particularly in data-limited settings.

Figures

Figures reproduced from arXiv: 2508.20513 by the authors.

Figure 1
Figure 1. The Overall Pipeline of MoTAS: MoE-Guided Feature Selection from TTS-Augmented Speech for Multimodal [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The Upper Part of the Diagram Illustrates the Framework of Our Proposed MoTAS; The Lower Part Details the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Test Accuracy on the ADReSSo Dataset [29] With Different TTS Augmentation Factors The results demonstrate that the MoE mechanism consistently enhances performance across settings. Without TTS augmentation, introducing MoE increased the test accuracy from 78.28% to 79.71% (ID1 vs. ID2), indicating its effectiveness under limited data con￾ditions. When applied to the 2× augmented dataset, MoE further improved accuracy… view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 32 canonical work pages

  1. [1]

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

    2020

  2. [2]

    Aparna Balagopalan and Jekaterina Novikova. 2021. Comparing acoustic-based approaches for Alzheimer’s disease detection. arXiv preprint arXiv:2106.01555 (2021)

    arXiv 2021

  3. [3]

    Subrato Bharati, Prajoy Podder, Dang Ngoc Hoang Thanh, and VB Surya Prasath

    work page

  4. [4]

    Jun Chen, Jieping Ye, Fengyi Tang, and Jiayu Zhou. 2021. Automatic detection of Alzheimer’s disease using spontaneous speech only. In Interspeech, Vol. 2021. 3830

    work page 2021

  5. [5]

    Xia Cui, Amila Gamage, Terry Hanley, and Tingting Mu. 2021. Identifying indicators of vulnerability from short speech segments using acoustic and textual features. Proceedings of Interspeech 2021 (2021), 1569–1573

    work page 2021

  6. [6]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers). 4171–4186

    2019

  7. [7]

    Kewen Ding, Madhu Chetty, Azadeh Noori Hoshyar, Tanusri Bhattacharya, and Britt Klein. 2024. Speech based detection of Alzheimer’s disease: a survey of AI techniques, datasets and challenges. Artificial Intelligence Review 57, 12 (2024), 325

    work page 2024

  8. [8]

    Cong-Thanh Do, Shuhei Imai, Rama Doddipatla, and Thomas Hain. 2024. Im- proving Accented Speech Recognition using Data Augmentation based on Un- supervised Text-to-Speech Synthesis. In 2024 32nd European Signal Processing Conference (EUSIPCO). IEEE, 136–140

    work page 2024

Show all 48 references
  1. [9]

    Elif Eyigoz, Sachin Mathur, Mar Santamaria, Guillermo Cecchi, and Melissa Naylor. 2020. Linguistic markers predict onset of Alzheimer’s disease. EClini- calMedicine 28 (2020)

    2020

  2. [10]

    William Fedus, Barret Zoph, and Noam Shazeer. 2022. Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity. Journal of Machine Learning Research 23, 120 (2022), 1–39

    2022

  3. [11]

    Kathleen C Fraser, Jed A Meltzer, and Frank Rudzicz. 2015. Linguistic features identify Alzheimer’s disease in narrative speech. Journal of Alzheimer’s disease 49, 2 (2015), 407–422

    2015

  4. [12]

    María Lara Gauder, Leonardo Daniel Pepino, Luciana Ferrer, and Pablo Riera. 2021. Alzheimer disease recognition using speech-based embeddings from pre-trained models. (2021)

    2021

  5. [13]

    Neeraj Gaur, Brian Farris, Parisa Haghani, Isabel Leal, Pedro J Moreno, Manasa Prasad, Bhuvana Ramabhadran, and Yun Zhu. 2021. Mixture of informed ex- perts for multilingual speech recognition. In ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal P...

    2021

  6. [14]

    H Goodglass, E Kaplan, and B Barresi. 2001. Boston Diagnostic Aphasia Exami- nation. Lippincott Williams & Wilkins. Philadelphia, PA (2001)

    2001

  7. [15]

    Gaurav Gupta, Meghana Kshirsagar, Ming Zhong, Shahrzad Gholami, and Juan Lavista Ferres. 2021. Comparing recurrent convolutional neural networks for large scale bird species classification. Scientific reports 11, 1 (2021), 17085

    2021

  8. [16]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition . 770–778

    2016

  9. [17]

    Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Rus- lan Salakhutdinov, and Abdelrahman Mohamed. 2021. Hubert: Self-supervised speech representation learning by masked prediction of hidden units. IEEE/ACM transactions on audio, speech, and language process...

    2021

  10. [18]

    Zhiheng Huang, Wei Xu, and Kai Yu. 2015. Bidirectional LSTM-CRF models for sequence tagging. arXiv preprint arXiv:1508.01991 (2015)

    2015 arXiv

  11. [19]

    Jonghwan Hyeon, Yung-Hwan Oh, Young-Jun Lee, and Ho-Jin Choi. 2024. Improv- ing speech emotion recognition by fusing self-supervised learning and spectral features via mixture of experts.Data & Knowledge Engineering 150 (2024), 102262

    2024

  12. [20]

    Yun-Hee Jeon, David Foxe, Guk-Hee Suh, Huali Wang, Jacqueline C Dominguez, Rex Maukera, Sengchanh Kounnavong, and Olivier Piguet. 2024. Post-diagnosis dementia care in the Western Pacific region: assessment of needs and pathways to optimal care. The Lancet Regional Health–West...

    2024

  13. [21]

    Arun Jha and Kaushik Mukhopadhaya. 2020. Alzheimer’s Disease: Diagnosis and Treatment Guide. Springer Nature

    2020

  14. [22]

    Jaehyeon Kim, Jungil Kong, and Juhee Son. 2021. Conditional variational autoen- coder with adversarial learning for end-to-end text-to-speech. In International Conference on Machine Learning . PMLR, 5530–5540

    2021

  15. [23]

    Siddique Latif, Abdullah Shahid, and Junaid Qadir. 2023. Generative emotional AI for speech emotion recognition: The case for synthetic emotional speech augmentation. Applied Acoustics 210 (2023), 109425

    2023

  16. [24]

    Jinpeng Li and Wei-Qiang Zhang. 2024. Whisper-based transfer learning for alzheimer disease classification: Leveraging speech segments with full transcripts as prompts. InICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE,...

    2024

  17. [25]

    Shijia Liao, Yuxuan Wang, Tianyu Li, Yifan Cheng, Ruoyi Zhang, Rongzhi Zhou, and Yijin Xing. 2024. Fish-speech: Leveraging large language models for advanced multilingual text-to-speech synthesis. arXiv preprint arXiv:2411.01156 (2024)

    2024 arXiv

  18. [26]

    Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Cheng- gang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, et al. 2024. Deepseek-v3 technical report. arXiv preprint arXiv:2412.19437 (2024)

    2024 arXiv

  19. [27]

    Xuan-Hao Liu, Wei-Bang Jiang, Wei-Long Zheng, and Bao-Liang Lu. 2024. MoGE: Mixture of Graph Experts for Cross-subject Emotion Recognition via Decompos- ing EEG. In 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). IEEE, 3515–3520

    2024

  20. [28]

    Saturnino Luz, Sofia de la Fuente, and Pierre Albert. 2018. A method for analysis of patient speech in dialogue for dementia detection.arXiv preprint arXiv:1811.09919 (2018)

    2018 arXiv

  21. [29]

    Saturnino Luz, Fasih Haider, Sofia De la Fuente, Davida Fromm, and Brian MacWhinney. 2021. Detecting cognitive decline using speech only: The adresso challenge. arXiv preprint arXiv:2104.09356 (2021)

    2021 arXiv

  22. [30]

    Pranav Mahajan and Veeky Baths. 2021. Acoustic and language based deep learn- ing approaches for Alzheimer’s dementia detection from spontaneous speech. Frontiers in Aging Neuroscience 13 (2021), 623607

    2021

  23. [31]

    PL McGeer, H Kamo, R Harrop, EG McGeer, WRW Martin, BD Pate, and DKB Li. 1986. Comparison of PET, MRI, and CT with pathology in a proven case of Alzheimer’s disease. Neurology 36, 12 (1986), 1569–1569

    1986

  24. [32]

    Bahman Mirheidari, Yilin Pan, Daniel Blackburn, Ronan O’Malley, and Heidi Christensen. 2021. Identifying Cognitive Impairment Using Sentence Represen- tation Vectors.. In Interspeech. 2941–2945

    2021

  25. [33]

    Yilin Pan, Bahman Mirheidari, Jennifer M Harris, Jennifer C Thompson, Matthew Jones, Julie S Snowden, Daniel Blackburn, and Heidi Christensen. 2021. Using the Outputs of Different Automatic Speech Recognition Paradigms for Acoustic-and BERT-Based Alzheimer’s Dementia Detection...

    2021

  26. [34]

    Elodie Passeri, Kamil Elkhoury, Margaretha Morsink, Kerensa Broersen, Michel Linder, Ali Tamayol, Catherine Malaplate, Frances T Yen, and Elmira Arab- Tehrany. 2022. Alzheimer’s disease: treatment strategies and their limitations. International journal of molecular sciences 23...

    2022

  27. [35]

    VM Praseetha and PP Joby. 2022. Speech emotion recognition using data aug- mentation. International Journal of Speech Technology 25, 4 (2022), 783–792

    2022

  28. [36]

    Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2023. Robust speech recognition via large-scale weak supervision. In International conference on machine learning . PMLR, 28492–28518

    2023

  29. [37]

    Ali N Salman, Karen Rosero, Lucas Goncalves, and Carlos Busso. 2025. Mixture of Emotion Dependent Experts: Facial Expressions Recognition in Videos through Stacked Expert Models. IEEE Open Journal of Signal Processing (2025)

    2025

  30. [38]

    Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. 2019. Dis- tilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter.arXiv preprint arXiv:1910.01108 (2019)

    2019 arXiv

  31. [39]

    Tian Tan, Yizhou Lu, Rao Ma, Sen Zhu, Jiaqi Guo, and Yanmin Qian. 2021. Aispeech-sjtu asr system for the accented english speech recognition challenge. In ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 6413–6417

    2021

  32. [40]

    Aaron Van Den Oord, Sander Dieleman, Heiga Zen, Karen Simonyan, Oriol Vinyals, Alex Graves, Nal Kalchbrenner, Andrew Senior, Koray Kavukcuoglu, et al

  33. [41]

    Ning Wang, Yupeng Cao, Shuai Hao, Zongru Shao, and KP Subbalakshmi. 2021. Modular Multi-Modal Attention Network for Alzheimer’s Disease Detection Using Patient Audio and Language Data.. In Interspeech. 3835–3839

    2021

  34. [42]

    Yuxuan Wang, RJ Skerry-Ryan, Daisy Stanton, Yonghui Wu, Ron J Weiss, Navdeep Jaitly, Zongheng Yang, Ying Xiao, Zhifeng Chen, Samy Bengio, et al . 2017. Tacotron: Towards end-to-end speech synthesis. arXiv preprint arXiv:1703.10135 (2017)

    2017 arXiv

  35. [43]

    Zhiyong Wang, Ruibo Fu, Zhengqi Wen, Jianhua Tao, Xiaopeng Wang, Yuankun Xie, Xin Qi, Shuchen Shi, Yi Lu, Yukun Liu, et al. 2025. Mixture of experts fusion for fake audio detection using frozen wav2vec 2.0. In ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech...

    2025

  36. [44]

    Guanrou Yang, Fan Yu, Ziyang Ma, Zhihao Du, Zhifu Gao, Shiliang Zhang, and Xie Chen. 2025. Enhancing Low-Resource ASR through Versatile TTS: Bridging the Data Gap. In ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 1–5

    2025

  37. [45]

    Qin Yang, Xin Li, Xinyun Ding, Feiyang Xu, and Zhenhua Ling. 2022. Deep learning-based speech analysis for Alzheimer’s disease detection: a literature review. Alzheimer’s Research & Therapy 14, 1 (2022), 186

    2022

  38. [46]

    Youxiang Zhu, Abdelrahman Obyat, Xiaohui Liang, John A Batsis, and Robert M Roth. 2021. Wavbert: Exploiting semantic and non-semantic speech using wav2vec and bert for dementia detection. In Interspeech, Vol. 2021. 3790

    2021

  39. [2016]

    Wavenet: A generative model for raw audio.arXiv preprint arXiv:1609.03499 12 (2016)

    2016 arXiv

  40. [2022]

    Multimedia Tools and Applications 81, 18 (2022), 25971–25992

    Dementia classification using MR imaging and clinical data with voting based machine learning models. Multimedia Tools and Applications 81, 18 (2022), 25971–25992

    2022

Pith tools

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