Neural Network based Activity Tracker

• Published in: 2018 19th International Conference on Research and Education in Mechatronics (REM) pp. 12 - 17
• Main Authors: Abu-Khalaf, Jumana, Bouri, Saleem El, Giha, Najib, Al-Chalabi, Lamya, Al-Halhouli, Alaaldeen
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2018
• Subjects: fitness tracking; Graphical user interfaces; health monitoring; Heart rate; Legged locomotion; Neural Network; Neural networks; Sensors; Temperature measurement; Training; wearable senor
• DOI: 10.1109/REM.2018.8421775

This paper describes the development of an artificial neural network that is able to predict simple human activities such as resting, walking, and running. A wearable sensor prototype that is able to measure a person's heart rate, blood oxygen saturation, body temperature and humidity was developed. Readings from the various sensors were used to train a neural network using MATLAB. Also in order to monitor all input data from the sensors and the network output a graphical user interface was built in MATLAB. The developed pulse oximeter sensor circuit gave relatively accurate readings in comparison to a commercial sensor when tested. The neural network resulted in accurate predictions of human activity even for human subjects that weren't part of the network's training. This system could be further developed to be used in applications such as health monitoring and fitness tracking.