Contact-based temperature, breathing and cough patterns dataset for early COVID-19 symptoms identification
No Thumbnail Available
Date
12/12/2021
Authors
Omer Ali
Mohamad Khairi Bhatti
Muhammad Kamran Liaquat
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Description
This release features the dataset recorded using our prototype hardware device (contact-based) that collects accelerometer and temperature readings to investigate breathing patterns, personal activity and cough patterns. As fever and cough are considered as two of the most common symptoms for COVID-19, our aim was to focus on these physiological features. In addition, research also states that all COVID-19 contractions showed elevated breathing patterns in patients, that could be easily identifiable from Eupnea state. Therefore, in this research project, we aimed at:
1. Designing a prototype chest-worn device to measure dynamic chest movements (to record breathing and cough patterns)
2. Detect different activity patterns
3. Record temperature variations during idle and active stage
4. Using unsupervised machine learning algorithm to detect anomalies by creating a composite score for (breathing and cough patterns as well as temperature).
This release also features some code examples that were used for data pre-processing, feature identification and anomaly detection using (K-means and DBSCAN algorithms).
It is important to note that this dataset should be considered and further investigated for preliminary exploratory analysis. The data was collected from healthy adults (that did not undergo COVID-19 clinical screening tests). Therefore, it must be clearly identified that this dataset DOES NOT represent positive COVID-19 contractions.
--Published Articles
[1]. O. Ali, M. K. Ishak and M. Kamran, "Early covid-19 symptoms identification using hybrid unsupervised machine learning techniques," Computers, Materials & Continua, vol. 69, no.1, pp. 747–766, 2021.
Keywords
covid-19, dataset, anomaly detection, feature extraction