Wireless heart-beat monitoring system with supervised learning / Lye Wei Liang... [et al.]

• Main Authors: Liang, Lye Wei, Yusuf Fadhlullah, Solahuddin, Abdullah, Samihah, Abdul Hamid, Shabinar
• Format: Article
• Language: English
• Published: Universiti Teknologi Mara Cawangan Pulau Pinang 2020
• Subjects: Pharmaceutical technology; Industrial research. Research and development
• Online Access: http://ir.uitm.edu.my/id/eprint/33233/1/33233.pdf; http://ir.uitm.edu.my/id/eprint/33233/

Most of the hospitals in Malaysia still utilise manual inspection by medical personnel to determine the health conditions of the patients. The data collected from the medical equipment would have to be analysed and verified by the hospital. Frequently, many patients need medical inspections. However, to provide a precise diagnosis, medical personnel requires more time. This limitation can be addressed by the development of automated and wireless health monitoring systems with health diagnostic feature supported by artificial intelligence (AI). In this project, the objective is to develop a prototype of a wireless (non-invasive) heartbeat monitoring system with supervised learning. This system monitors the heartbeat activity and predicts the condition of the user's heartbeat. Technically, a photoplethysmography-based (PPG-based) heartbeat sensor is used to build a heartbeat sensing device with a Bluetooth feature that communicates with an Android application. The Android application is developed to receive heartbeat data from the device and feed the data into an AI classification model to predict the heartbeat condition of the user. This AI classifier was built from heartbeat data collected from 10 healthy people. The additional heartbeat dataset was generated based on a sound source of heartbeat information to increase the volume of the training dataset. The completion of this project implementation results in a wireless heartbeat monitoring system that can be applied regardless of location and time. The accuracy of the AI prediction is 99 % when evaluated with a testing dataset. The empirical accuracy obtained by testing the system with actual implementation is 90 %.