Novel Deep Learning-Based Vocal Biomarkers for Stress Detection in Koreans.

Area of Science:

• Digital Health
• Biomedical Engineering
• Psychology

Background:

• Societal changes necessitate effective stress detection and management.
• Chronic mental stress is linked to physical and psychological illnesses.
• Individuals often lack awareness of their stress levels, necessitating regular monitoring.

Purpose of the Study:

• Investigate the effectiveness of vocal biomarkers for stress detection in healthy Korean employees.
• Contribute to the development of digital healthcare solutions for stress management.
• Assess the potential of voice analysis in identifying stress indicators.

Main Methods:

• A multi-center clinical study involved 115 healthy Korean employees.
• Voice recordings were collected under relaxed and stress-induced conditions (socially evaluated cold pressor test).
• The Emphasized Channel Attention, Propagation and Aggregation in Time delay neural network (ECAPA-TDNN) architecture was used for stress prediction.

Main Results:

• The developed model achieved a 70% accuracy rate in stress detection.
• Vocal biomarkers demonstrated potential as a convenient tool for stress self-monitoring.
• The findings support the integration of voice analysis into digital healthcare frameworks.

Conclusions:

• Voice-based stress assessment shows promise within the Korean population.
• Continued research on vocal biomarkers across diverse linguistic groups is important.
• Vocal biomarkers can aid in proactive stress management and digital health interventions.