An open-source audio database for respiratory rate (RR) estimation research
Explore the docs »
View Metadata
Table of Contents
Respiratory rate is a well-known acoustic biomarker of the health status of the respiratory system.
In this repository, we offer a systematic and labelled set of auditory recordings of 31 (15 male, 16 female) subjects, aged between 18 and 25 years-old, breathing at five different respiratory rates, organized by breaths per minute (bpm), captured with two different recording hardware.
Each recording takes 1 minute, and was registered at 4500 Hz, one channel, 32-bit resolution.
Adsitional information about the health history of each subject is also provided, along with some other demographic data as heigh and weight.
The main objective of the "Respiratory Rate universities of Jaén and Oviedo" (RRuJO) dataset is to provide a set of reliably labeled sound signals at different respiratory rates so that they can be used by researchers to evaluate different algorithms applied to respiratory rate estimation from the analysis of sound respiratory signals.
The database is organized in folders according to the following structure:
- RRinervasO: recordings with the designed stethoscope (ESP32 based, Littman bell, INMP441 microphone)
- Folder named '8': recordings at 8 bpm
- Folder named '10': recordings at 10 bpm
- Folder named '12': recordings at 12 bpm
- Folder named '18': recordings at 18 bpm
- Folder named '20': recordings at 20 bpm
- ThinklabsO: recordings with the commercial Thinklabs stethoscope
- Folder named '8': recordings at 8 bpm
- Folder named '10': recordings at 10 bpm
- Folder named '12': recordings at 12 bpm
- Folder named '18': recordings at 18 bpm
- Folder named '20': recordings at 20 bpm
Inside each subfolder you will find 31 recordings following the CODE column of the Metadata.xlsx file.
Example. This figure shows an example of the signal XXX.wav, belonging to the RRinervasO set, with a RR=10.
Figure 1
The database is published along with an Excel file with the information about the subjects recorded. You can see this information in a friendly mode here. In that file, you will find the following fields:
- Code: Temporal signature of the recording by the following format: year (four digits), month (two digits), day (two digits), hour (two digits), minutes (two digits) and the research group that captured the recording (University of Jaén=1, University of Oviedo=2). Thus, as an example, a participant auscultating at the University of Jaén on February 5, 2023 at 17:42 (format 24h) would obtain the following code: 2023020517421.
- Age (in years)
- Sex (M = Male, F = Female)
- Smoker? (0 = Non-smoker, 1 = Smoker)
- Heigh (in cm)
- Weight in Kg)
- Status? (0 = Healthy, 1 = Sick)
- Pathology: indicates what disease was the subject suffering from while being auscultated.
Distributed under the MIT License. See LICENSE.txt for more information.
José Ranilla Pastor, QHPC group - [email protected]
Group Link: [https://pirweb.edv.uniovi.es/] (QHPC Group)
Project Link: https://github.com/QHPC-SP-Research-Lab/Respiratory-Rate-Database/
When using this dataset please cite the following publication “An orthogonal non-negative matrix factorization approach for respiratory rate estimation using a wireless stethoscope” as the source.
This database has been supported in part under grant PID2020-119082RB-{C21,C22} funded by MCIN/AEI/10.13039/501100011033 of the Ministerio de Ciencia e Innovación de España, and grant AYUD/2021/50994 funded by Gobierno del Principado de Asturias, Spain.