“We wanted to make continuous heart rate variability measurements in real-world environments easier to study—and easier to use in future health tools.”
— Sungkyu Shaun Park

Wearable health monitors have become everyday companions—tracking steps, sleep, and sometimes stress. But can they offer something more scientifically useful? In this study, researchers present the first open-access dataset of continuous, real-world heart rate variability (HRV) collected via smartwatches—spanning four weeks of daily life for 49 healthy individuals.

The result is a rich, multi-modal resource that blends raw physiological signals with daily sleep logs and validated clinical assessments of anxiety, depression, and insomnia. Unlike previous HRV datasets collected in controlled lab environments, this one captures the complexity and variability of actual human routines—offering a benchmark for developing AI-driven tools that assess mental well-being in the wild.

Why sleep and mood data matter together

Heart rate variability is a subtle but powerful signal of autonomic nervous system function. It’s known to decrease with chronic stress, disrupted sleep, and poor mental health. That’s why the researchers didn’t just record HRV—they also paired it with daily sleep diaries and three rounds of clinical surveys, including the PHQ-9, GAD-7, and ISI.

This combination enables a more holistic view of mental state over time. Participants reported their sleep and symptoms while passively generating over 33,000 hours of wearable data. The dataset captures circadian trends, moment-to-moment variability, and the kinds of lifestyle irregularities that traditional studies miss. In doing so, it supports the development of predictive analytics for mental health.

Making wearable data usable for science

Each participant wore a Samsung Galaxy Watch Active 2, which collected photoplethysmography (PPG) signals at 100ms intervals—fast enough to compute HRV in 5-minute windows. Alongside motion and light data, the researchers derived key time- and frequency-domain HRV features (e.g. SDNN, RMSSD, LF/HF) and flagged signal quality with a “missingness score” for every chunk.

The team validated the data by showing expected HRV rhythms across time of day, sex, and age, and by confirming correlations with movement and sleep. The final release includes raw PPG data, computed HRV features, activity metrics, daily sleep logs, and three rounds of psychological survey data—all openly available under a Creative Commons license.

How the researchers tested it

After an initial onboarding session, participants wore the device for four continuous weeks—charging it at night and syncing data via Wi-Fi. Each day, they logged sleep onset and wake time, while clinical surveys were administered at the start, midpoint, and end of the study. The research team monitored compliance in real time and provided reminders through a private group chat.

To compute HRV, the team used the open-source HeartPy Python library, which handles noisy PPG signals common in free-living data. Only high-quality 5-minute segments were retained for analysis. The result is a uniquely clean, continuous, and ecologically valid dataset for studying short-term HRV in everyday life.

“We’re not just measuring heartbeats — we’re observing behavior, stress, and recovery in the wild.”

What it means for future mental health monitoring

As interest in wearable-based mental health tracking grows, this dataset lays important groundwork. It offers not only a methodological template, but also a new standard for the types of data needed to build interpretable, real-time wellness models. Researchers can use it to benchmark detection systems for sleep irregularity, anxiety spikes, or burnout risk.

Most importantly, this work shows that it’s possible to gather clinically relevant biosignals outside of a lab—at scale, with open tools, and in a way that preserves privacy. It invites a shift from snapshot diagnostics to continuous understanding, where mental health support can be proactive, not reactive.

About the Paper

Title: A continuous real-world dataset comprising wearable-based heart rate variability alongside sleep diaries
Authors: Aitolkyn Dulatova, Sungkyu Shaun Park, Marios Constantinides, Sangwon Lee, Daniele Quercia, Meeyoung (Mia) Cha
Publication status: Published in Scientific Data, 2025 (DOI: 10.1038/s41597-025-05801-3)
Data: Available on Figshare at doi.org/10.6084/m9.figshare.28509740
Design highlights: 49 participants, 4-week smartwatch study, 100ms PPG sampling, 5-minute HRV segments, daily sleep logs, tri-weekly clinical surveys, open-source release.