Stress Detection from Photoplethysmography Signals Using Multi-Domain Heart Rate Variability Analysis

Abstract

Stress is a key contributor to cardiovascular and mental health disorders, which motivates reliable, non intrusive monitoring methods for early detection and intervention. Photoplethysmography (PPG) has emerged as a practical alternative to electrocardiography (ECG) for wearable cardiovascular monitoring because of its low cost and ease of integration into consumer devices. This study presents a stress detection framework based on multi domain pulse rate variability (PRV) analysis derived from PPG signals. Publicly available datasets, including the Wearable Stress and Affect Detection (WESAD) dataset and a Pulse Transit Time (PTT) PPG dataset, are used to validate the proposed signal processing and feature extraction pipeline. Raw PPG signals are preprocessed with bandpass filtering and robust peak detection to obtain inter beat intervals, from which time domain, frequency domain, and nonlinear PRV features, as well as the Baevsky Stress Index, are computed. On WESAD, statistically significant differences between baseline and TSST induced stress conditions are observed in key PRV features (e.g., SDNN, RMSSD, pNN50, LF/HF ratio, SD1, SD2, ApEn, SampEn; Bonferroni corrected p < 0.0029), indicating reduced variability, a shift toward sympathetic dominance, and decreased dynamical complexity during stress. A subset of features also shows strong stability across window lengths from 2 to 5 minutes (R² ≥ 0.6 when compared to 5 minute reference segments), supporting their use in near real time monitoring. These results confirm the feasibility of PPG based, multi domain PRV analysis for non invasive stress assessment and highlight its suitability for wearable and edge based healthcare applications.