Pith. sign in

REVIEW 4 major objections 7 minor 40 references

Handcrafted acoustic features from raw spontaneous speech alone can screen for Alzheimer's under strict speaker-independent evaluation, with a lightweight SVM reaching mean AUC 0.674.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-14 13:46 UTC pith:PSLEX2ZM

load-bearing objection Honest, modest audio-only baseline with real evaluation hygiene; useful reference numbers, not a method advance. the 4 major comments →

arxiv 2607.10168 v1 pith:PSLEX2ZM submitted 2026-07-11 cs.SD cs.AI

Transcript-Free Lightweight Detection of Alzheimer's Disease from Spontaneous Speech Using Handcrafted MFCC-Dominant Acoustic Biomarkers

Rashin Gholijani Farahani , Azam Bastanfard This is my paper
classification cs.SD cs.AI
keywords Alzheimer’s diseasespontaneous speechacoustic biomarkerspause analysisWebRTC VADMFCCtranscript-freespeaker-independent evaluation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper sets out to show that Alzheimer's disease can be detected from spontaneous speech using only audio, without transcripts, ASR, or heavy deep models. From 176 balanced Cookie Theft recordings, WebRTC voice-activity detection separates speech from silence; 99 handcrafted features (pause and fluency statistics, spectral-prosodic descriptors, and MFCC summaries with first- and second-order deltas) are extracted and fed to a simple RBF-SVM. Across 30 speaker-independent GroupShuffleSplit runs the mean AUC is 0.674, while pause-only features fall near or below chance and MFCC-based descriptors carry most of the signal. A sympathetic reader cares because this supplies a transparent, CPU-only, language-agnostic baseline that could support low-resource or privacy-preserving digital triage when neuroimaging or language tools are unavailable.

Core claim

Transcript-free spectro-temporal and fluency-related cues extracted from raw audio facilitate speaker-independent Alzheimer's screening: a lightweight RBF-SVM on 99 handcrafted features (VAD pauses, spectral-prosodic summaries, and MFCC+Δ+ΔΔ) attains mean AUC 0.674 ± 0.091 across 30 GroupShuffleSplit iterations on the balanced 176-recording Pitt Cookie Theft set, establishing a practical deployment-oriented baseline.

What carries the argument

The 99-dimensional handcrafted feature vector—13 WebRTC-VAD pause/fluency statistics, 8 acoustic-prosodic summaries (ZCR, spectral centroid/bandwidth/flux), and 78 MFCC+Δ+ΔΔ mean/std descriptors—inside a leakage-safe Pipeline of mean imputation, standard scaling, and RBF-SVM trained only on training speakers.

Load-bearing premise

The chosen voice-activity detector and handcrafted statistics mainly capture disease-related speech changes rather than recording artifacts or speaker-specific silence habits.

What would settle it

Re-running the identical 99-feature RBF-SVM under the same speaker-independent protocol on an independent spontaneous-speech corpus with controlled recording conditions and matched demographics yields AUC near chance (≈0.5).

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Transcript-free, CPU-only acoustic screening is feasible with classical models rather than ASR-dependent or deep pipelines.
  • MFCC-based spectro-temporal descriptors, not pause statistics alone, drive stable speaker-independent discrimination.
  • A compact subset of roughly 20 features can retain much of the signal, provided selection is nested inside each training split.
  • The pipeline supplies a reproducible, leakage-aware baseline for deployment-oriented audio-only research.
  • Language-independent operation is theoretically supported, though cross-corpus validation remains necessary.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar lightweight audio filters could serve as first-pass triage before expensive multimodal or imaging work-ups in primary-care or telehealth settings.
  • Because pause-only features collapse under speaker independence, VAD aggressiveness and pause thresholds should be treated as tunable hyperparameters rather than fixed constants.
  • The gap between group-level pause separation and predictive failure suggests recording-condition artifacts may dominate small spontaneous-speech corpora more than previously assumed.
  • Nested feature selection plus external validation on larger multi-task, multi-language cohorts is the natural next test of whether the reported AUC generalizes.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes a lightweight, transcript-free baseline for Alzheimer’s disease (AD) detection from spontaneous speech using 176 balanced Cookie Theft recordings (88 AD / 88 HC) from the DementiaBank Pitt corpus. After WebRTC VAD segmentation, 99 handcrafted features are extracted (13 pause/fluency statistics, 8 spectral–prosodic descriptors, and 78 MFCC+Δ+ΔΔ mean/std summaries) and classified with an RBF-SVM under strict speaker-independent GroupShuffleSplit evaluation over 30 splits, yielding mean AUC 0.674 ± 0.091 (illustrative single-split AUC 0.742). Feature-group ablations show pause-only features below chance (AUC 0.432), while MFCC-only (0.654) and the full set (0.674) carry the signal; an exploratory non-nested Top-20 subset (AUC 0.719) is explicitly flagged as potentially optimistic. The authors conclude that spectro-temporal and fluency-related acoustic cues can support speaker-independent, audio-only AD screening as a deployment-oriented baseline.

Significance. If the reported signal is genuinely AD-related rather than residual recording confounds, the work supplies a useful, reproducible, CPU-only audio baseline that prioritizes speaker independence, interpretability, and language independence—properties often missing from transcript/ASR or heavy deep models. Strengths include the repeated GroupShuffleSplit protocol with mean±std, the transparent feature-group ablation (Table IV), explicit non-nested selection caveats for the Top-20 experiment, and honest positioning as a modest baseline rather than a state-of-the-art claim. These practices raise the bar for small-corpus speech-AD papers and give the community a clear reference point for transcript-free pipelines. The clinical significance remains limited by the modest AUC, single-task N=176 design, and unresolved channel/artifact risk, so the main contribution is methodological and baseline-setting rather than diagnostic readiness.

major comments (4)
  1. Section IV-F / Table IV and Limitations (V-D, citing Gauder et al. [37]): the primary claim that the AUC reflects AD-related spectro-temporal/fluency biomarkers is load-bearing but under-supported. Ablation shows pause-only AUC 0.432 (below chance) while MFCC-only reaches 0.654 and full-99 0.674, so nearly all discrimination is MFCC-dominant. Speaker-independent splits block identity leakage but not systematic channel/condition differences that may correlate with labels (clinic, session, mic, encoding). The paper itself flags heterogeneous recording artifacts as a known risk on DementiaBank-style data, yet provides no artifact-control experiment (e.g., silence/non-speech-only classification, channel-matched subsets, or noise/codec ablation). Without such a control or substantially stronger caveats in Abstract/Conclusion, the interpretation that the 0.674 AUC is cognitive/articulatory rat
  2. Abstract and Section IV-H / Conclusion: the repeated phrasing that “transcript-free spectro-temporal and fluency-related cues” facilitate screening overstates the fluency component relative to the evidence. Table IV shows pause-only features fail under the same speaker-independent protocol that is used for the main claim; descriptive pause shifts (Figs. 1–2) do not translate into predictive utility. The abstract and summary should be revised so that primary conclusions rest on MFCC/spectro-temporal signal, with pause/fluency features described as descriptively interesting but not standalone discriminators in this setting.
  3. Section III-B (VAD settings) and free parameters: WebRTC aggressiveness=2, 30 ms frames, and the 0.2 s pause threshold (plus short/long cutoffs 0.3 s / 1.0 s) are fixed without sensitivity analysis, yet pause statistics—and to a lesser extent speech-mask-dependent spectral summaries—depend on them. Given that pause-only performance collapses under SI evaluation, a brief sensitivity or alternative-VAD check (or explicit statement that main conclusions are MFCC-driven and VAD-threshold-robust) is needed so readers can judge whether the pipeline’s temporal branch is reproducible or brittle.
  4. Dataset scope (III-A, V-D, VI): all results rest on N=176 from a single elicitation task (Cookie Theft) with binary labels and no external cohort. The paper acknowledges this, but the central “practical foundation for deployment-oriented research” claim still needs either (i) a clearer delimitation that the contribution is an in-corpus SI baseline only, or (ii) at least one external/cross-task check or power discussion. As written, generalization language in Abstract/Conclusion slightly exceeds what the design can support.
minor comments (7)
  1. Abstract reports “average AUC of 0.674” without the ±0.091 given in the body; align abstract and Table II for consistency.
  2. Throughout (e.g., Abstract, I, III): English is often non-idiomatic (“We take out 99…”, “documenting performance across 30 iterations”, “The makeup of the dataset…”). A careful language edit would improve clarity without changing content.
  3. Table V: useful contextualization, but several rows mix Acc/F1/AUC and different protocols; a short column note or footnote restating non-comparability (already in text) next to the table would help skimmers.
  4. Figs. 1–3: captions describe distributions well, but axis labels/units and sample sizes per class in the figure panels themselves would aid readability in print.
  5. Section III-D: XGBoost is listed as “optional” yet appears in Table II; clarify whether it is a primary baseline or supplementary.
  6. Keywords and title emphasize “MFCC-Dominant,” which matches Table IV; ensure Abstract lead sentence does not foreground fluency equally with spectro-temporal cues after revision.
  7. References: a few entries appear duplicated or near-duplicate (e.g., HAFFormer arXiv vs ICASSP forms [33]/[34]); clean bibliography.

Circularity Check

1 steps flagged

No load-bearing circularity; standard speaker-independent supervised baseline with one acknowledged non-nested exploratory subset that is excluded from primary claims.

specific steps
  1. fitted input called prediction [Section III-F / IV-E (Exploratory Feature-Subset Analysis)]
    "In the current implementation, feature importance is computed once on the full dataset and the same Top-20 subset is reused across splits. Because selection is not nested within each training split, this analysis may introduce optimistic bias and is therefore treated as exploratory rather than a primary result. ... SVM AUC: 0.719 ± 0.091"

    Ranking on the entire corpus (including future test speakers) then reusing the fixed Top-20 for all splits statistically inflates the reported compact-subset AUC relative to a properly nested procedure; the authors correctly flag and demote it, so it does not contaminate the primary full-99 claim.

full rationale

The paper reports an empirical audio-only classification pipeline (WebRTC VAD + 99 handcrafted pause/spectral/MFCC features + RBF-SVM) evaluated under repeated GroupShuffleSplit speaker-independent splits on the balanced Pitt Cookie Theft set. Primary performance (mean AUC 0.674 ± 0.091) is obtained by training and scaling exclusively on training speakers and scoring held-out speakers; nothing reduces by construction to a fitted constant or self-defined quantity. Feature-group ablations and the single-split illustration are likewise ordinary held-out measurements. The sole minor issue is the exploratory Top-20 subset, whose Random-Forest ranking is performed once on the full dataset (not nested); the authors explicitly label it optimistic, exclude it from primary conclusions, and note the proper nested alternative. No self-definitional equations, no fitted-input-as-prediction for the main result, no load-bearing self-citation uniqueness claim, and no renaming of a known result as a derivation. The work is therefore self-contained against its own evaluation protocol; residual concerns about recording artifacts are correctness/confound issues, not circularity.

Axiom & Free-Parameter Ledger

4 free parameters · 3 axioms · 0 invented entities

The paper is an empirical ML baseline; its load-bearing content rests on standard signal-processing assumptions, a handful of hand-chosen thresholds, and the claim that the extracted features carry AD-related information under speaker-independent evaluation. No new physical entities are postulated. Free parameters are the usual VAD and pause cut-offs plus default classifier settings; domain axioms are the usual ones linking temporal/spectral speech changes to cognitive decline.

free parameters (4)
  • WebRTC VAD aggressiveness = 2
    Set to 2 by hand; controls speech/non-speech frame decisions that define all pause features.
  • pause duration threshold = 0.2 s (count), 0.3/1.0 s (ratios)
    Pauses shorter than 0.2 s are discarded; short/long ratios use 0.3 s and 1.0 s cut-offs. These are conventional but arbitrary choices that directly shape the 13 temporal features.
  • SVM C and gamma = C=1.0, gamma=scale
    Defaults C=1.0, gamma='scale' are used without nested hyper-parameter search; they affect the reported AUC surface.
  • GroupShuffleSplit test_size and n_splits = 0.2 / 30
    test_size=0.2, n_splits=30, random_state=42 fix the evaluation distribution whose mean AUC is the central number.
axioms (3)
  • domain assumption Temporal pause statistics and MFCC spectro-temporal summaries extracted from spontaneous speech carry information about Alzheimer’s-related cognitive-linguistic decline that is partially independent of lexical content.
    Invoked throughout Introduction and Related Works; underpins the decision to discard transcripts entirely.
  • domain assumption Speaker identity can be blocked by grouping on participant_id extracted from filenames and using GroupShuffleSplit, yielding an unbiased estimate of generalization to new speakers.
    Section III-A; the entire evaluation protocol rests on this grouping being correct and complete.
  • ad hoc to paper WebRTC VAD at the chosen settings produces speech/non-speech masks sufficiently accurate for pause statistics to be meaningful on DementiaBank Cookie Theft audio.
    Section III-B; no calibration or comparison to human pause annotation is provided.

pith-pipeline@v1.1.0-grok45 · 19042 in / 3077 out tokens · 34568 ms · 2026-07-14T13:46:06.288355+00:00 · methodology

0 comments
read the original abstract

It is still hard to find Alzheimer's disease (AD) early, especially when neuroimaging is expensive or tools that depend on language are not available. Spontaneous speech provides a non-invasive signal; however, numerous current methodologies depend on transcripts/ASR or computationally intensive deep models. We offer a simple, audio-only baseline for detecting AD using 176 Cookie Theft recordings from the DementiaBank Pitt corpus (88 AD, 88 controls). WebRTC voice activity detection (VAD) is used to separate speech from non-speech. We take out 99 hand-crafted acoustic-temporal features, including pause and fluency statistics, spectral/prosodic descriptors, and MFCC summaries with {\Delta} and {\Delta}{\Delta}. Evaluation is performed using a stringent speaker-independent GroupShuffleSplit,documenting performance across 30 iterations. A lightweight SVM with an RBF kernel gets an average AUC of 0.674 across runs. For example, a single split has an AUC of 0.742 and an accuracy of 0.657. We also present an exploratory compact-feature analysis utilizing a Top-20 subset ranked by Random Forest importance; since selection is not nested within training splits, these results may be overly optimistic and are not employed for primary conclusions (AUC 0.719). The results indicate that transcript-free spectro-temporal and fluency-related cues can facilitate speaker-independent Alzheimer's disease screening from raw audio, establishing a practical foundation for deployment-oriented research.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 30 canonical work pages · 1 internal anchor

  1. [1]

    Machine learning approaches for speech-based Alzheimer’s detection: A comprehensive survey,

    A. Sharafeldeen, J. Keowen, and A. Shaffie, “Machine learning approaches for speech-based Alzheimer’s detection: A comprehensive survey,” Computers, vol. 14, no. 2, p. 36, 2025, doi: 10.3390/computers14020036

    work page doi:10.3390/computers14020036 2025

  2. [2]

    Noninvasive automatic detection of Alzheimer’s disease from spontaneous speech: A review,

    X. Qi, Q. Zhou, J. Dong, and W. Bao, “Noninvasive automatic detection of Alzheimer’s disease from spontaneous speech: A review,” Frontiers in Aging Neuroscience, vol. 15, 2023, Art. no. 1224723, doi: 10.3389/fnagi.2023.1224723

    work page doi:10.3389/fnagi.2023.1224723 2023

  3. [3]

    Artificial intelligence, speech, and language processing approaches to monitoring Alzheimer’s disease: A systematic review,

    S. de la Fuente Garcia, C. W. Ritchie, and S. Luz, “Artificial intelligence, speech, and language processing approaches to monitoring Alzheimer’s disease: A systematic review,” Journal of Alzheimer’s Disease , vol. 78, no. 4, pp. 1547 –1574, 2020, doi: 10.3233/JAD-200888

    work page doi:10.3233/jad-200888 2020

  4. [4]

    Voiceprints of cognitive impairment: Analyzing digital voice for early detection of Alzheimer’s and related dementias,

    N. Rezaii et al ., “Voiceprints of cognitive impairment: Analyzing digital voice for early detection of Alzheimer’s and related dementias,” npj Dementia, vol. 1, p. 35, 2025, doi: 10.1038/s44400-025-00040-0

    work page doi:10.1038/s44400-025-00040-0 2025

  5. [5]

    Pauses for detection of Alzheimer’s disease,

    J. Yuan, X. Cai, Y. Bian, Z. Ye, and K. Church, “Pauses for detection of Alzheimer’s disease,” Frontiers in Computer Science, vol. 2, 2021, Art. no. 624488, doi: 10.3389/fcomp.2020.624488

    work page doi:10.3389/fcomp.2020.624488 2021

  6. [6]

    Speech based detection of Alzheimer’s disease: A survey of AI techniques, datasets and challenges,

    K. Ding, M. Chetty, A. N. Hoshyar, T. Bhattacharya, and B. Klein, “Speech based detection of Alzheimer’s disease: A survey of AI techniques, datasets and challenges,” Artificial Intelligence Review , vol. 57, p. 325, 2024, doi: 10.1007/s10462-024-10961-6

    work page doi:10.1007/s10462-024-10961-6 2024

  7. [7]

    Telltale silence: Temporal speech parameters discriminate between prodromal dementia and mild Alzheimer’s disease,

    V. Vincze et al ., “Telltale silence: Temporal speech parameters discriminate between prodromal dementia and mild Alzheimer’s disease,” Journal of Communication Disorders, vol. 89, 2020, Art. no. 105978

    2020

  8. [8]

    Alzheimer’s disease classification from speech pause distributions with context information,

    G. Khatri, R. Soleimani, K. L. Haley, A. Jacks, and E. Lobaton, “Alzheimer’s disease classification from speech pause distributions with context information,” in Proc. 46th Annu. Int. Conf. IEEE Engineering in Medicine and Biology Society (EMBC), 2024, pp. 1–6, doi: 10.1109/EMBC53108.2024.10782834

    doi:10.1109/embc53108.2024.10782834 2024

  9. [9]

    An automated approach to examining pausing in the speech of people with dementia,

    R. A. Sluis et al., “An automated approach to examining pausing in the speech of people with dementia,” American Journal of Alzheimer’s Disease & Other Dementias , vol. 35, pp. 1 –8, 2020, doi: 10.1177/1533317520939773

    work page doi:10.1177/1533317520939773 2020

  10. [10]

    Efficient pause extraction and encode strategy for Alzheimer’s disease detection using only acoustic features from spontaneous speech,

    J. Liu et al ., “Efficient pause extraction and encode strategy for Alzheimer’s disease detection using only acoustic features from spontaneous speech,” Brain Sciences, vol. 13, no. 3, p. 477, 2023, doi: 10.3390/brainsci13030477

    work page doi:10.3390/brainsci13030477 2023

  11. [11]

    Voice quality and speech fluency distinguish individuals with mild cognitive impairment from healthy controls,

    C. Themistocleous, M. Eckerström, and D. Kokkinakis, “Voice quality and speech fluency distinguish individuals with mild cognitive impairment from healthy controls,” PLOS ONE, vol. 15, no. 7, 2020, Art. no. e0236009, doi: 10.1371/journal.pone.0236009

    work page doi:10.1371/journal.pone.0236009 2020

  12. [12]

    Exploring high spectro-temporal resolution for Alzheimer’s dementia detection,

    N. A. Vats, P. Barche, G. S. Mirishkar, and A. K. Vuppala, “Exploring high spectro-temporal resolution for Alzheimer’s dementia detection,” in Proc. IEEE Int. Conf. on Signal Processing and Communications (SPCOM), 2022, pp. 1–5, doi: 10.1109/SPCOM55316.2022.9840847

    work page doi:10.1109/spcom55316.2022.9840847 2022

  13. [13]

    Alzheimer’s Disease Detection from Speech Using Shapley Additive Explanations for Feature Selection and Enhanced Interpretability,

    I. Oiza -Zapata and A. Gallardo -Antolín, "Alzheimer’s Disease Detection from Speech Using Shapley Additive Explanations for Feature Selection and Enhanced Interpretability," Electronics, vol. 14, no. 11, p. 2248, 2025, doi: 10.3390/electronics14112248

    work page doi:10.3390/electronics14112248 2025

  14. [14]

    Deep learning of speech data for early detection of Alzheimer’s disease in the elderly,

    K. Ahn et al ., “Deep learning of speech data for early detection of Alzheimer’s disease in the elderly,” Bioengineering, vol. 10, no. 9, p. 1093, 2023, doi: 10.3390/bioengineering10091093

    work page doi:10.3390/bioengineering10091093 2023

  15. [15]

    Integrating Pause Information with Word Embeddings in Language Models for Alzheimer’s Disease Detection from Spontaneous Speech,

    Y. Pu and W. -Q. Zhang, “Integrating Pause Information with Word Embeddings in Language Models for Alzheimer’s Disease Detection from Spontaneous Speech,” in Proc. ICASSP 2025 – IEEE Int. Conf. Acoustics, Speech and Signal Processing (ICASSP), 2025, pp. 1–5, doi: 10.1109/ICASSP49660.2025.10888563

    doi:10.1109/icassp49660.2025.10888563 2025

  16. [16]

    Alzheimer’s disease detection using deep learning and machine learning: A review,

    S. Mohsen, “Alzheimer’s disease detection using deep learning and machine learning: A review,” Artificial Intelligence Review, vol. 58, p. 262, 2025, doi: 10.1007/s10462-025-11258-y

    work page doi:10.1007/s10462-025-11258-y 2025

  17. [17]

    A review of machine learning approaches for non-invasive cognitive impairment detection,

    M. Alsuhaibani et al., “A review of machine learning approaches for non-invasive cognitive impairment detection,” IEEE Access, vol. 13, pp. 56355–56384, 2025, doi: 10.1109/ACCESS.2025.3555176

    work page doi:10.1109/access.2025.3555176 2025

  18. [18]

    Machine -learning-based disease diagnosis: A comprehensive review,

    M. M. Ahsan, S. A. Luna, and Z. Siddique, “Machine -learning-based disease diagnosis: A comprehensive review,” Healthcare, vol. 10, no. 3, p. 541, 2022, doi: 10.3390/healthcare10030541

    work page doi:10.3390/healthcare10030541 2022

  19. [19]

    An explainable machine learning approach for Alzheimer’s disease classification,

    A. S. Alatrany, W. Khan, A. Hussain, H. Kolivand, and D. Al-Jumeily, “An explainable machine learning approach for Alzheimer’s disease classification,” Scientific Reports , vol. 14, p. 2637, 2024, doi: 10.1038/s41598-024-51985-w

    work page doi:10.1038/s41598-024-51985-w 2024

  20. [20]

    Early detection of Alzheimer’s disease using deep learning methods,

    A. C. Mmadumbu, F. Saeed, F. Ghaleb, and S. N. Qasem, “Early detection of Alzheimer’s disease using deep learning methods,” Alzheimer’s & Dementia , vol. 21, p. e70175, 2025, doi: 10.1002/alz.70175

    work page doi:10.1002/alz.70175 2025

  21. [21]

    Early detection and classification of Alzheimer’s disease through data fusion of MRI and DTI images using the YOLOv11 neural network,

    W. Hechkel and A. Helali, “Early detection and classification of Alzheimer’s disease through data fusion of MRI and DTI images using the YOLOv11 neural network,” Frontiers in Neuroscience , vol. 19, Art. no. 1554015, 2025, doi: 10.3389/fnins.2025.1554015

    work page doi:10.3389/fnins.2025.1554015 2025

  22. [22]

    Clinical validation of digital healthcare solutions: State of the art, challenges and opportunities,

    M. Gomis -Pastor et al ., “Clinical validation of digital healthcare solutions: State of the art, challenges and opportunities,” Healthcare, vol. 12, no. 11, p. 1057, 2024, doi: 10.3390/healthcare12111057

    work page doi:10.3390/healthcare12111057 2024

  23. [23]

    The Persian linguistic based audio-visual data corpus, AVA II, considering coarticulation,

    A. Bastanfard, M. Fazel, A. A. Kelishami, and M. Aghaahmadi, “The Persian linguistic based audio-visual data corpus, AVA II, considering coarticulation,” in MMM 2010, LNCS 5916, S. Boll et al., Eds. Berlin, Germany: Springer, 2010, pp. 284 –294, doi: 10.1007/978 -3-642- 11301-7_30

    doi:10.1007/978 2010

  24. [24]

    Automatic analysis of linguistic features in journal articles of different academic impacts with feature engineering techniques,

    S. Lei, R. Yang, and C. -R. Huang, “Automatic analysis of linguistic features in journal articles of different academic impacts with feature engineering techniques,” arXiv preprint, arXiv:2111.07525, 2021

    Pith/arXiv arXiv 2021

  25. [25]

    Generalization gap in data augmentation: Insights from illumination,

    J. Xiao, W. Guo, J. Liu, and M. Li, “Generalization gap in data augmentation: Insights from illumination,” arXiv preprint , arXiv:2404.07514v3, 2024

    Pith/arXiv arXiv 2024

  26. [26]

    OBGESS: Automating original Bender Gestalt Test based on one stage deep learning,

    M. Fathi Ahmadsaraei, A. Bastanfard, and A. Amini, “OBGESS: Automating original Bender Gestalt Test based on one stage deep learning,” International Journal of Computational Intelligence Systems, vol. 16, p. 178, 2023, doi: 10.1007/s44196-023-00353-z

    work page doi:10.1007/s44196-023-00353-z 2023

  27. [27]

    High-performance medicine: the convergence of human and artificial intelligence,

    A. Topol, “High-performance medicine: the convergence of human and artificial intelligence,” Nature Medicine, vol. 25, no. 1, pp. 44 –56, 2019, doi: 10.1038/s41591-018-0300-7

    doi:10.1038/s41591-018-0300-7 2019

  28. [28]

    Unveiling the sound of the cognitive status: Machine learning -based speech analysis in the Alzheimer’s disease spectrum,

    F. García-Gutiérrez et al., “Unveiling the sound of the cognitive status: Machine learning -based speech analysis in the Alzheimer’s disease spectrum,” Alzheimer’s Research & Therapy, vol. 16, p. 26, 2024, doi: 10.1186/s13195-024-01394-y

    work page doi:10.1186/s13195-024-01394-y 2024

  29. [29]

    Preclinical Alzheimer’s disease: Pathophysiology, biomarker staging, and implications for early detection,

    C. R. Jack et al., “Preclinical Alzheimer’s disease: Pathophysiology, biomarker staging, and implications for early detection,” medRxiv, Jan. 2023, doi: 10.1101/2023.01.26.23285064

    work page doi:10.1101/2023.01.26.23285064 2023

  30. [30]

    The natural history of Alzheimer’s disease: Description of study cohort and accuracy of diagnosis,

    J. T. Becker, F. Boller, O. L. Lopez, J. Saxton, and K. L. McGonigle, “The natural history of Alzheimer’s disease: Description of study cohort and accuracy of diagnosis,” Archives of Neurology , vol. 51, no. 6, pp. 585 –594, 1994, doi: 10.1001/archneur.1994.00540180063015

    work page doi:10.1001/archneur.1994.00540180063015 1994

  31. [31]

    Acoustic Speech Analysis in Alzheimer’s Disease: A Systematic Review and Meta-Analysis,

    S. Saeedi, S. Hetjens, M. O. W. Grimm, and B. Barsties V. Latoszek, “Acoustic Speech Analysis in Alzheimer’s Disease: A Systematic Review and Meta-Analysis,” J. Prev. Alzheimers Dis., vol. 11, no. –, pp. –, 2024, doi: 10.14283/jpad.2024.132

    work page doi:10.14283/jpad.2024.132 2024

  32. [32]

    The impact of pause and filler word encoding on dementia detection with contrastive learning,

    R. Soleimani, S. Guo, K. L. Haley, A. Jacks, and E. Lobaton, “The impact of pause and filler word encoding on dementia detection with contrastive learning,” Applied Sciences, vol. 14, no. 19, p. 8879, Oct. 2024, doi: 10.3390/app14198879

    work page doi:10.3390/app14198879 2024

  33. [33]

    HAFFormer: A hierarchical attention-free framework for Alzheimer’s disease detection from spontaneous speech,

    Z. Dong, Z. Zhang, W. Xu, J. Han, J. Ou, and B. W. Schuller, “HAFFormer: A hierarchical attention-free framework for Alzheimer’s disease detection from spontaneous speech,” arXiv:2405.03952 [cs.SD], May 2024

    Pith/arXiv arXiv 2024

  34. [34]

    HAFFormer: A hierarchical attention-free framework for Alzheimer’s disease detection from spontaneous speech,

    Z. Dong, Z. Zhang, W. Xu, J. Han, J. Ou, and B. W. Schuller, “HAFFormer: A hierarchical attention-free framework for Alzheimer’s disease detection from spontaneous speech,” in Proc. IEEE Int. Conf. Acoustics, Speech and Signal Processing (ICASSP), 2024, pp. 11246– 11250, doi: 10.1109/ICASSP48485.2024.10446795

    doi:10.1109/icassp48485.2024.10446795 2024

  35. [35]

    Alzheimer's Disease Detection from Spontaneous Speech and Text: A review

    V. M. K., G. V., A. P. R., D. S., and S. K. G., “Alzheimer’s disease detection from spontaneous speech and text: A review,” arXiv preprint, arXiv:2307.10005, 2023, doi: 10.48550/arXiv.2307.10005

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2307.10005 2023

  36. [36]

    Detection of Mild Cognitive Impairment from Non - Semantic, Acoustic Voice Features: The Framingham Heart Study,

    K. Ding et al., “Detection of Mild Cognitive Impairment from Non - Semantic, Acoustic Voice Features: The Framingham Heart Study,” JMIR Aging, 2024, e55126, doi: 10.2196/55126

    work page doi:10.2196/55126 2024

  37. [37]

    The Unreliability of Acoustic Systems in Alzheimer’s Speech Datasets with Heterogeneous Recording Conditions,

    L. Gauder et al., "The Unreliability of Acoustic Systems in Alzheimer’s Speech Datasets with Heterogeneous Recording Conditions," arXiv:2409.12170v1 [cs.SD], Sep. 2024

    Pith/arXiv arXiv 2024

  38. [38]

    Automated bimodal pause analysis for acoustic markers of cognitive decline and Alzheimer’s disease in connected speech,

    D. He et al., "Automated bimodal pause analysis for acoustic markers of cognitive decline and Alzheimer’s disease in connected speech," Alzheimer’s & Dementia, vol. 21, p. e70635, 2025, doi: 10.1002/alz.70635

    work page doi:10.1002/alz.70635 2025

  39. [39]

    Harnessing acoustic speech parameters to decipher amyloid status in individuals with mild cognitive impairment,

    F. García-Gutiérrez et al., "Harnessing acoustic speech parameters to decipher amyloid status in individuals with mild cognitive impairment," Frontiers in Neuroscience, vol. 17, Art. no. 1221401, 2023, doi: 10.3389/fnins.2023.1221401

    work page doi:10.3389/fnins.2023.1221401 2023

  40. [40]

    A comprehensive audio -visual corpus for teaching sound Persian phoneme articulation,

    A. Bastanfard, A. A. Kelishami, M. Fazel and M. Aghaahmadi, "A comprehensive audio -visual corpus for teaching sound Persian phoneme articulation," 2009 IEEE International Conference on Systems, Man and Cybernetics, San Antonio, TX, USA, 2009, pp. 169- 174, doi: 10.1109/ICSMC.2009.5346591

    work page doi:10.1109/icsmc.2009.5346591 2009