Development of machine learning techniques for non-invasive heart rate sensor

• Main Author: Lohesh Gnanasekaran
• Other Authors: School of Electrical and Electronic Engineering
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/136595
• Institution: Nanyang Technological University

It has been recorded that in Singapore, an average of 17 deaths per day were due to Cardiovascular Diseases (CVD) and it was also the top cause of death in the world, according to the World Health Organization (WHO). One of the most basic ways to detect them would be fluctuating heart rates. Both Bradycardia and Tachycardia, which are the occurrence of low and fast heart rates respectively, are good indicators of the presence of some CVDs. There is therefore a need for a low cost and non-invasive portable heart rate monitoring systems where heart rates can be monitored and be informative. In the current market, there are numerous ways to measure heart rate however there are very minimal devices that are focused only for health application. Sensors and infrared have been installed into wearable devices that provide readings of heart rate and blood pressure, however reading accuracy varies and majority of smart watches focuses on non-health application. The accuracy and functionality of these readings are important for those who are medically required to monitor and record their heart rates consistently. The aim of this project is to test the accuracy and the reliability of a low-budget unbranded smartwatch in comparison to a hospital-approved Omron heart rate monitor. Heart rate readings will be taken from both Omron and the unbranded Smart Watch and will be classified into three categories: Bradycardia, Tachycardia and Normal. This data will be stored into the WEKA machine learning software where different types of classifiers will be applied. The best classifier that provides the most accurate result will be chosen and modeled using python. The classifier will be compared to both the Python model and to WEKA in terms of mean, median, standard deviation and accuracy result. Overall, Random Forest classifier has proven to be a better classifier and can be easily adapted in smaller devices such as Smart Watches for better functionality towards heart rate monitoring.