Enhanced Resolution Analysis of Photoplethysmography Signals for Stress Monitoring Using Taylor-Fourier Method

Heart rate variability (HRV) is widely recognized as an indicator of autonomic nervous system activity and stress responses. Photoplethysmography (PPG) provides a non-invasive means of estimating HRV, but traditional peak-detection methods are susceptible to motion artifacts and irregular pulse morphologies, limiting their reliability. This study investigates the Taylor-Fourier (TF) method as an alternative approach, modeling PPG signals as quasi-periodic waveforms with time-varying amplitude and phase. By estimating instantaneous heart rate (IHR) with higher temporal resolution, the TF method provides a continuous representation of heart rate dynamics that may capture additional stress-related variations. The proposed pipeline includes TF-based IHR estimation, HRV feature extraction, and classification using machine learning to compare its performance with conventional IBI-based methods. We apply this approach to the publicly available WESAD dataset, which includes both baseline and stress conditions. While both methods produce comparable time-domain HRV estimates, TF-based frequency-domain metrics exhibit notable differences. To assess its applicability in stress detection, we train machine learning models on HRV features extracted using both methods. Results indicate that TF-derived features can improve classification performance in some models, though subject-specific variability remains a factor. These findings, while still preliminary, suggest that the TF approach could enhance HRV-based stress monitoring, warranting further exploration of its robustness in real-world applications.