Stress detection using time-frequency analysis and machine learning framework.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Psychophysiology

Background:

• Chronic stress negatively impacts physical and mental health, leading to conditions like depression and heart disease.
• Early stress detection is crucial for mitigating adverse health effects.
• Physiological signals offer objective measures for stress assessment.

Purpose of the Study:

• To develop and validate a Machine Learning (ML) based model for accurate and efficient stress identification.
• To explore the utility of Inter Beat Interval (IBI) and Electro Dermal Activity (EDA) signals for stress detection.
• To reduce computational requirements for stress identification models.

Main Methods:

• Utilized K-EmoCon and WESAD datasets, acquiring IBI and EDA signals via Empatica E4 wristband.
• Extracted Time-Frequency features using Ensemble Empirical Mode Decomposition (EEMD) and Hilbert Transform (HT).
• Employed traditional ML models, including k-Nearest Neighbors (k-NN), with Instantaneous Frequency (IF) as input.

Main Results:

• The k-NN model achieved high performance, with an accuracy of approximately 99.85% and an F1-score of 99.87%.
• The proposed approach demonstrated reduced computational power requirements, suitable for resource-limited environments.
• Validation with real-time data from a Fitbit smartwatch confirmed the model's efficiency and improved performance.

Conclusions:

• The ML-based model effectively identifies stress using IBI and EDA signals with high accuracy.
• The method offers a computationally efficient approach for real-time stress monitoring.
• This research contributes to proactive mental and physical health management through early stress detection.