A Pillow-Based Ballistocardiography and ECG Dataset with Manual J-Peak Annotations for Sleep Monitoring

PillowBCG-ECG_JPeaks is a publicly available dataset for ballistocardiography (BCG) heartbeat analysis collected in a natural sleep environment. The dataset was acquired using a bed-based, pillow-mounted piezoelectric sensor for BCG measurement, with simultaneously recorded electrocardiography (ECG) signals serving as a reference. Data were collected from five healthy subjects over eight overnight sessions.All recordings are segmented into fixed-length epochs of 30 seconds. For each epoch, BCG J-peak locations were manually annotated with guidance from the synchronized ECG signals. The dataset provides both discrete J-peak indices and sample-level binary masks aligned with the BCG waveform, enabling direct use in classical signal processing pipelines as well as deep learning\u2013based segmentation models.The dataset is fully anonymized, with subjects and nights identified using non-identifiable indices. It is intended to support research on BCG heartbeat detection, BCG\u2013ECG relationship modeling, and unobtrusive cardiovascular monitoring during sleep. By providing manually annotated BCG J-peaks recorded in an unconstrained sleep setting, this dataset aims to facilitate reproducible benchmarking and methodological development in biomedical signal processing.