A Dataset for Quantifying Sleep Quality from Wristband Pulse Data via the ESD Model

Nighttime wristband pulse data were collected from 28 subjects using a commercial smart band (Keep Smart Band, Keep B3), with a sampling interval of 15 seconds and a maximum recording duration of 8 hours, covering complete sleep episodes. The dataset comprises 549 sleep samples across three categories of subjects: (1) one adult male with abnormal sleep patterns due to night shifts (22 samples); (2) three healthy adults with long-term recordings (73 samples); and (3) 24 clinical patients receiving six-week sleep intervention treatment at Peking University First Hospital (454 samples).The dataset provides raw pulse rate time series, computed local pulse rate variability (Δ), sleep efficiency function (W), effective sleep duration (ESD), and daily subjective sleep scores (DSS) for each subject.

The ESD model establishes a linear relationship between objective pulse-derived metrics and subjective sleep scores (DSS = k·ESD + α), achieving high consistency between objective and subjective sleep quality assessments (overall correlation coefficient r ≈ 0.92).This dataset can be used for the development and validation of sleep quality assessment algorithms, improvement of wearable-device-based sleep monitoring methods, and objective evaluation of clinical sleep intervention outcomes.