Related Experiment Video
Updated: May 31, 2026
05:48
Memorization-Based Training and Testing Paradigm for Robust Vocal Identity Recognition in Expressive Speech Using Event-Related Potentials Analysis
Published on: August 9, 2024
Psychosocial Stress in the Chinese Community: Speech Analytics Through Linguistic and Acoustic Fusion Using Machine
Amanda M Y Chu1, Benson S Y Lam2, Jenny T Y Tsang3
1Department of Social Sciences and Policy Studies, The Education University of Hong Kong, Tai Po, Hong Kong, China.
JMIR Biomedical Engineering
|May 29, 2026
Summary
Detecting family caregiver stress is vital. A new machine learning model fuses linguistic and acoustic speech features, achieving 78.28% AUC, to effectively identify psychosocial stress.
Area of Science:
- Psychology
- Computer Science
- Health Informatics
Background:
- Family caregivers face significant stress, increasing susceptibility to adverse psychosocial health conditions.
- Early detection of caregiver stress is crucial for timely interventions and preventing long-term disability.
Purpose of the Study:
- To develop and validate a machine learning approach for psychosocial stress assessment in family caregivers.
- To enhance stress assessment effectiveness through fusion analysis of linguistic and acoustic speech features.
Main Methods:
- Quantitative analysis of speech data from 100 Chinese family caregivers.
- Extraction of linguistic and acoustic speech features and application of machine learning classifiers (e.g., support vector machine).
- Orthogonalization procedure to decorrelate features before fusion analysis.
Main Results:
- The linear support vector machine model achieved 78.28% AUC, 75.27% F1-score, and 73% accuracy.
- Fusion of linguistic and acoustic features significantly outperformed models using single feature types.
- Orthogonalization of features before fusion enhanced classification accuracy.
Conclusions:
- Fusion analysis of linguistic and acoustic features effectively identifies psychosocial stress in family caregivers.
- Proper feature processing is essential for combining multiple speech features.
- Findings support developing machine learning models for psychosocial stress assessment and population mental health management.
