A Mixed Cloud-and-Embedded-based Approach with Machine Learning Towards the Development of a Fall Monitoring System

• Main Authors: Limbaga, Neil Joshua P., Mallari, Kevin Luis T., Yeung, Nathan Richward O., Oppus, Carlos M
• Format: text
• Published: Archīum Ateneo 2023
• Subjects: 3-axis Accelerometer; Cloud Computing; Edge Impulse Computing; Fall Monitoring System; Full Stack Development; Shallow Neural Network; Computer Sciences; Electrical and Computer Engineering; Engineering; Physical Sciences and Mathematics
• Online Access: https://archium.ateneo.edu/ecce-faculty-pubs/154; https://doi.org/10.1109/ECTIDAMTNCON57770.2023.10139738
• Institution: Ateneo De Manila University

The ability to monitor falls, especially for the elderly, deems to be a crucial task to provide quality and timely healthcare response. However, there have been minimal efforts in centralizing such activity for efficient hospital management. This paper presents the development of a full-stack fall monitoring system with edge computing and machine learning technologies. Using a 3-axis accelerometer of a smartphone, motion data is collected and directly sent to an edge computing platform wherein a shallow neural network is directly trained to classify the motion data into positional states: stable, falling sidewards, falling flat, and standing up. A confusion matrix is presented to evaluate the performance of the neural network model, both in training and in real time. A cloud-based approach using ReactJS for front-end integration and Firebase's Cloud Firestore with NodeJS embedded capabilities for real-time data storage and embedded classification is implemented.