Psychological stress level detection based on electrodermal activity.

Area of Science:

• Physiological signal processing
• Psychological stress detection
• Machine learning for health

Background:

• Current stress detection methods often rely on multiple physiological signals, increasing complexity.
• There is a need for practical, single-signal-based approaches for stress monitoring.

Purpose of the Study:

• To investigate the feasibility of using electrodermal activity (EDA) as a single physiological signal for detecting psychological stress levels.
• To develop and evaluate a stress detection method based on EDA and linear discriminant analysis (LDA).

Main Methods:

• Feature extraction from electrodermal activity (EDA) signals across different driving conditions (rest, highway, city driving).
• Application of Fisher projection and Linear Discriminant Analysis (LDA) for stress level classification (low, medium, high).
• Utilized cross-validation strategies including leave-one-out and test cross-validation.

Main Results:

• Achieved an 81.82% recognition rate for classifying stress levels using EDA and LDA.
• The single-signal approach demonstrated a favorable balance between recognition performance and computational load compared to multi-signal systems.

Conclusions:

• The proposed EDA-based method offers a practical and computationally efficient alternative for psychological stress detection.
• This research presents a promising avenue for developing practical personal stress monitoring devices.