Differentiation of stress and physical activity in wearable ECG signals using fuzzy entropy metrics

Stress detection using Electrocardiography (ECG) signals are considered as a direct measure of the autonomic nervous system activity, becoming increasingly sensitive to stress-induced physiological changes. Heart Rate Variability features extracted from the ECG signal may result in a similar physiological response for both stress and physical activity. Alternatively, entropy-based metrics capture inherent morphological dynamics and can be used to detect stress from ECG signals. This study employs fuzzy entropy features to distinguish stress from rest (baseline) and physical activity conditions in real time. The ECG signals are collected from 36 participants using a data acquisition system. The acquired signals are preprocessed and normalized to remove noises, artifacts and heart rate effects. Further, it is down sampled to 250Hz to reduce computational complexity for real time analysis and fuzzy entropy features are extracted. Results show that fuzzy entropy exhibits reduced value in stress condition compared to the baseline which may be due to increased sympathetic activity and also indicates statistically significant (p<0.05) results between them. Further, analyzing the time window for effective discrimination of stress and physical activity reveals that 5s window demonstrates consistent distinctions between them, and can be considered for real-time stress analysis. Thereby, the fuzzy entropy feature can be used for effective discrimination of stress from physical activity using ECG signals.