To investigate the effect of R-R interval (RRI) normalization using different scaling factors ( \({\sigma }^{n}\) ) on frequency-domain heart rate variability (HRV) parameters during sleep, and determine whether normalization improves the differentiation of sleep stages and enhances the accuracy of slow-wave sleep (SWS) detection. We analyzed the overnight electrocardiogram data of 130 participants. RRIs were normalized using a proposed z-score method with scaling factors (n = 1–4 in \({\sigma }^{n}\) ). Frequency-domain HRV parameters were extracted using raw and normalized RRIs within a 5-min time window shifted by 30 s. HRV parameters were statistically compared across sleep stages (wakefulness, light sleep, SWS, and rapid eye movement sleep), and the SWS detection performance was evaluated using receiver operating characteristic analysis. RRI normalization significantly magnified differences in HRV parameters across sleep stages. \(HF\) , which reflects parasympathetic activity, showed a markedly clearer separation between SWS and other stages with normalization (p < 0.05) than with the raw RRI. Consequently, the SWS detection performance was significantly improved. The area under the curve for \(HF\) -based SWS detection increased from 0.50 (raw RRI) to a maximum of 0.86 (with \({\sigma }^{2}\) normalization). HRV parameters derived from normalized RRI reveal the hierarchical organization of autonomic nervous activity across sleep stages, particularly parasympathetic dominance in SWS. This method significantly improves SWS detection accuracy and provides a robust foundation for refined HRV-based sleep monitoring systems. It can be utilized in real-time health monitoring systems to accurately assess sleep stages, stress, and fatigue levels.
