Speech Features as Mental Health Biomarkers: A Systematic Analysis
A new study from arXiv (2605.24678) systematically analyzes perceptual speech features—prosody, vocal quality, semantic coherence, syntactic structure, and sarcasm—as objective cues for mental health assessment. Using interpretable machine learning (XGBoost with SHAP and LIME), researchers examined associations between speech characteristics and validated symptom measures of depression, anxiety, and ADHD. The framework was evaluated on controlled benchmarks (StressID, DAIC-WOZ, Androids, EATD) and a real-world clinical dataset. Results show stable relationships between symptom severity and vocal irregularities (shimmer, jitter), lexical-syntactic patterns, and affective tone. An ablation study identified the most informative features across datasets, supporting clinical decision-support applications.
Key facts
- Study uses perceptual speech features including prosody, vocal quality, semantic coherence, syntactic structure, and sarcasm.
- Machine learning models XGBoost with SHAP and LIME were employed.
- Evaluated on datasets: StressID, DAIC-WOZ, Androids, EATD, and a real-world clinical dataset.
- Features associated with depression, anxiety, and ADHD symptom measures.
- Vocal irregularities (shimmer, jitter) show consistent relationships with symptom severity.
- Lexical-syntactic patterns and affective tone are also significant.
- An ablation study identified the most informative features.
- The framework aims to support clinical decision-making in mental health care.
Entities
Institutions
- arXiv