BioSARdBScene: A Synthetic Biomedical Auscultation Scene Dataset (reproducible build specification)

A synthetic biomedical auscultation scene dataset for organ identification and normal/abnormal screening under realistic clinical noise, distributed as a reproducible build specification rather than as audio. Each item is a mixture in which a target body sound, such as heart, lung, bowel or knee joint, is combined with a second physiological sound from a different organ, a recorded hospital ambient scene, and background speech, each layer added at an individually measured signal-to-noise ratio and the secondary organ passed through a chest-wall transmission filter. This is the acoustic situation a digital stethoscope meets at the bedside rather than the isolated, quiet recording that public organ-sound corpora provide. 25,289 mixtures · 139.7 hours · 4 organs × 2 conditions This record contains the complete per-layer provenance and measured SNR of every layer of every mixture, the canonical group-disjoint split, a derived cepstral feature set, a SHA-256 manifest of the original audio, and the tool that rebuilds the corpus. It does not contain the mixture audio. Source corpora you need to download Rebuilding requires nine public corpora, obtained directly from their publishers under their own terms. Roughly 33,441 distinct source files, about 25 GB. DOWNLOAD_SOURCES.md gives the directory layout the rebuild expects and per-corpus notes on how to split normal from abnormal; rebuild_biosardbscene.py --check-sources reports anything missing before it writes a single file. Corpus Feeds Files Licence Download COUGHVID (Orlandic et al.) Lung primary and secondary 16,224 CC BY 4.0 10.5281/zenodo.4048312 PhysioNet/CinC Challenge 2016 (Liu et al.) Heart primary and secondary 3,153 ODC-BY 1.0 physionet.org/content/challenge-2016 CirCor DigiScope (Oliveira et al.) Heart primary and secondary 3,120 ODC-BY 1.0 10.13026/tshs-mw03 Trustworthy Intent in Speech (Maltezou-Papastylianou et al.) Speech layer, every mixture 1,151 CC BY 4.0 10.17605/OSF.IO/45D8J ICBHI 2017 Respiratory Sound Database (Rocha et al.) Lung primary and secondary 920 Research use; no formal redistribution terms bhichallenge.med.auth.gr BMD-HS (Ali et al., BUET) Heart primary and secondary 863 CC BY-NC-ND 4.0 github.com/sani002/BMD-HS-Dataset Bowel sounds (Ficek et al.) Bowel primary 590 CC BY-NC 4.0 10.34740/kaggle/dsv/2789758 Hospital Ambient Noise (Ali et al., BUET) Ambient layer, every mixture 562 CC BY-NC-ND 4.0 kaggle.com/datasets/nafin59/hospital-ambient-noise Lung sounds, electronic stethoscope (Fraiwan et al.) Lung primary and secondary 336 CC BY 4.0 10.17632/jwyy9np4gv.3 Knee-joint vibroarthrography (Krishnan) Joint primary 83 Author-held github.com/srikrishnan1972/Biomedical-Signal-Analysis