TiLoT 4.0 - Industry 4.0 base v-Community Platform (Fall Detection System)

Voon, Qi Aun (2022) TiLoT 4.0 - Industry 4.0 base v-Community Platform (Fall Detection System). Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text Full Text - Voon Qi Aun.pdf Restricted to Registered users only Download (3MB)

Abstract

Fall is scientifically defined as a downward motion from a higher to a lower position under the force of gravity due to a sudden loss or lack of support or balance. According to World Health Organization (WHO), falls are, globally, the second most prevalent factor of unintentional injury and death. There are approximately 28–35% of people with 65 year of age and above fall each year and the likelihood of fall increases significantly to 32–42% for people ageing over 70 years. Therefore, an automated fall detection which automatically identifies fall event in real time and sends alarm to emergency contact promptly is an essential system for shortening the waiting time for the fallen person. The two main approaches used detecting fall is ambient based devices (sensor) or vision-based input devices such as cameras. This paper proposes a sensor-based detection system using moving average and thresholding as the preliminary fall detection. If the data passes the preliminary detection, then a support vector machine model is applied for final detection. Smartwatch which consists of built-in MEMS sensors, including accelerometer, gyroscope is deployed to collect motion data in real time. When SVM model is tested with online dataset, it gives an accuracy of 1.0 and recall of 0.99 and 0.98 for F1-score. The fall detection algorithm is complimented with a communication platform which allows users (help-seeker) to interact with (helper) by sharing fall data conveniently. On top of that, if fall is detected and no response reply is received for some time or help request is sent by user, then smartphone will send a SMS message and make a phone call to user’s emergency contact to receive immediate help for the user