Datasetpublic

abmallick/heart-breath-sleep-stage-dataset

Heart‑Breath Sleep Stage Dataset A public dataset for physiological sleep analysis that contains heart rate and respiratory rate signals alongside sleep stage labels, collected over overnight sleep sessions. Each row represents a short time window from a night recording with derived cardiorespiratory features and the corresponding sleep stage. Dataset link: https://huggingface.co/datasets/abmallick/heart-breath-sleep-stage-dataset 📊 Dataset Overview Name:… See the full description on the dataset page: https://huggingface.co/datasets/abmallick/heart-breath-sleep-stage-dataset.

sourceHugging Facemitupdated 10mo agoView on Hugging Face

2likes36downloads

Dataset Card

Heart‑Breath Sleep Stage Dataset

A public dataset for physiological sleep analysis that contains heart rate and respiratory rate signals alongside sleep stage labels, collected over overnight sleep sessions. Each row represents a short time window from a night recording with derived cardiorespiratory features and the corresponding sleep stage.

Dataset link: https://huggingface.co/datasets/abmallick/heart-breath-sleep-stage-dataset

📊 Dataset Overview

—Name: Heart‑Breath Sleep Stage Dataset
—Author: Abhinav Mallick
—Modalities: Tabular / Time‑series (derived features)
—Format: Parquet
—License: MIT

🧠 Motivation

Sleep staging is traditionally performed using EEG‑based polysomnography, which is expensive, intrusive, and difficult to scale. This dataset focuses on heart rate (HR) and respiratory rate (RR) dynamics, which are known to correlate strongly with sleep stages, enabling:

—Non‑EEG sleep staging research
—Wearable and mattress‑based sleep monitoring
—Low‑cost, scalable sleep health solutions

The dataset is suitable for both classical ML and deep learning approaches.

📋 Features / Columns

Each row corresponds to a time window within a sleep recording.

Column	Description
`time_seconds`	Time since start of recording (seconds)
`minutes_since_start`	Time since start (minutes)
`heart_rate`	Instantaneous heart rate (bpm)
`respiratory_rate`	Instantaneous respiratory rate
`hr_mean`	Mean heart rate for the window
`hr_sdnn_5`	SDNN HRV metric (5‑min window)
`hr_rmssd_5`	RMSSD HRV metric (5‑min window)
`hr_slope_3`	Heart‑rate trend (3‑min slope)
`rr_mean`	Mean respiratory rate
`rr_sd_5`	Respiratory rate standard deviation
`rr_slope_3`	Respiratory‑rate trend
`hr_rr_ratio`	HR to RR ratio
`hr_rr_product`	HR × RR derived feature
`sleep_stage`	Sleep stage label (integer‑encoded)
`classes`	Human‑readable sleep stage label
`night_id`	Unique identifier per night/session
`split`	Dataset split (`train` / `val` / `test`)

🧩 Sleep Stage Labels

The sleep_stage column is integer‑encoded. Depending on usage, labels can be mapped to:

—Wake
—REM
—Light sleep (N1/N2)
—Deep sleep (N3)

Users may collapse or remap classes based on their modeling strategy.

🚀 Quick Start

Installation

bash

pip install datasets pandas pyarrow

Load Dataset

python

from datasets import load_dataset

dataset = load_dataset(
    "abmallick/heart-breath-sleep-stage-dataset",
    split="train"
)

print(dataset[0])

Convert to Pandas

python

df = dataset.to_pandas()
df.head()

📈 Example Use Cases

—Sleep stage classification models
—HRV and respiratory pattern analysis across sleep cycles
—Time‑series modeling with RNNs, TCNs, or Transformers
—Research on non‑EEG sleep monitoring systems
—Feature engineering for wearable or edge‑AI pipelines

🧪 Suggested Evaluation Metrics

—Accuracy
—Macro / Weighted F1‑score
—Cohen’s Kappa
—Per‑class Recall

📚 Citation

If you use this dataset in your research or projects, please cite:

bibtex

@misc{mallick2025heartbreathsleep,
  title={Heart‑Breath Sleep Stage Dataset},
  author={Mallick, Abhinav},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/abmallick/heart-breath-sleep-stage-dataset}
}

🤝 Contributions

Issues, improvements, and discussions are welcome via the Hugging Face dataset page.