Wearable technology is a promising tool for everyday health monitoring, with heart rate variability (HRV) providing key insights into current and potential health conditions. However, previous HRV datasets were collected under controlled clinical conditions, rather than in complex real-world environments. Here, we collected continuous physiological and motion signals using smartwatches from 49 healthy individuals (mean age: 28.35 ± 5.87, 51% females) over four weeks. The recordings were sampled every 100 ms, allowing for short-term HRV computation based on 5-minute segments of raw data. We validated the data by examining the frequency of collected signals, analyzing the correlation between the smartwatch sensor data and computed HRV, and demonstrating the presence of HRV and sleep-related feature distributions expected from the literature. Our wearable recordings were collected alongside daily self-reported sleep diaries and biweekly clinical questionnaires that assessed anxiety, depression, and insomnia. The dataset aims to benchmark in-the-wild HRV recordings, enable future analyses in the field, and support the development of predictive analytics that use sleep patterns and wearable data as health indicators.
Baigutanova et al. (Sat,) studied this question.