Deep Transfer Learning-Based Fall Detection Approach Using IoMT-Enabled Thermal Imaging-Assisted Pervasive Surveillance and Big Health Data

Abstract

People’s need for healthcare capacity has become increasingly critical as the elderly population continues to grow in most communities. Approximately 25–47% of seniors fall annually, and early detection of poor balance can significantly reduce their risk. Automated fall detection with big data analytics is key to maintaining the safety of the elderly in smart cities. Visible image systems (VIS) in smart buildings, on the other hand, visible image systems (VIS) in smart buildings may compromise the privacy of seniors by enabling technologies for intelligent big data analytics (IBDA). Thermal imaging (TI) is less obtrusive than visual imaging and can be used in combination with machine vision to perform a wide range of IBDAs. In this study, we present a novel two-step method for detecting falls in TI frames using deep learning (DL). As the first step, tracking tools are used to locate people’s locations. A novel modified deep transfer learning (TL) technique is used to classify the trajectory created by the tracking approach for people who are at risk of falling. Fall detection by the IBDA will be connected to the Internet of medical things (IoMT) and used as smart technology in the process of big data-assisted pervasive surveillance and health analytics. According to an analysis of the publicly available thermal fall dataset, our method outperforms traditional fall detection methods, with an average error of less than 3%. Additionally, IoMT platforms facilitate data processing, real-time monitoring and healthcare management. Our smart scheme for using big data analytics to enable intelligent decisions is compatible with the various spaces and provides a comfortable and safe environment for current and future elderly people.