What question did this study set out to answer?

To evaluate whether supervised machine learning models can accurately distinguish autistic from neurotypical adults using vowel acoustic features.

August 15, 2026Research in autism spectrum disordersOpen Access

Machine learning classification of autistic adults using vowel acoustic features

View Full Paper

Ask AI

Bookmark

To evaluate whether supervised machine learning models can accurately distinguish autistic from neurotypical adults using vowel acoustic features.
Extracted acoustic features—including fundamental frequency (F0), formants (F1–F3), duration, jitter, shimmer, harmonics-to-noise ratio (HNR), and intensity—from controlled vowel production tasks.
Trained and evaluated four supervised machine learning models (LightGBM, Random Forest, Support Vector Machine, and XGBoost) using token-level splitting and speaker-independent cross-validation.
Identified feature importance and model interpretability using SHAP (SHapley Additive exPlanations) analyses.
Under token-level validation, models achieved strong classification performance, with the best-performing model reaching an AUC of approximately 0.89.
Classification performance decreased substantially under speaker-independent cross-validation to AUCs of approximately 0.60 with wider confidence intervals.
SHAP analyses demonstrated that fundamental frequency (F0) was the strongest predictive feature across validation schemes, followed by sound intensity.

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Cross-sectional machine learning study reveals acoustic vowel markers distinguish autistic adults under token-level evaluation, suggesting generalization challenges across unseen speakers.

Cite This Study

Georgiou et al. (2026) studied this question.

synapsesocial.com/papers/6a80197375c2e31742c85757 https://doi.org/10.1016/j.reia.2026.202995

View Full Paper

Ask AI

Bookmark

Implication

Cross-sectional machine learning study reveals acoustic vowel markers distinguish autistic adults under token-level evaluation, suggesting generalization challenges across unseen speakers.

Machine learning classification of autistic adults using vowel acoustic features

Key Points

Discussion

Implication

Cite This Study

Machine learning classification of autistic adults using vowel acoustic features

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider