A systematic evaluation of window lengths for photoplethysmography-based sleep staging

Abstract

Sleep is vital for maintaining physiological functions and mental health. While photoplethysmography (PPG) from wearable devices offers a feasible solution for home-based sleep monitoring, existing methods primarily rely on overnight data acqui- sition, limiting their applicability in real-time sleep interventions. This study presents a three-class sleep staging framework based on short-term PPG signals, systematically investigating the staging performance across five different window lengths (3.5–50.5 min). Both 10.5-minute and 25.5-minute windows achieve effective sleep staging, with the 25.5-minute window delivering the highest Kappa among the evaluated window lengths (Kappa = 0.812). The shorter 10.5-minute window is suitable for rapid sleep quality assessment and daytime nap monitoring, while the 25.5-minute window is recommended for nocturnal sleep analysis and clinical screening, where higher stability is preferred for nocturnal sleep analysis and clinical screening. This work establishes a technical foundation for implementing timely sleep interventions on wearable devices, promoting the effective integration of health wearables into connected healthcare systems.