Recognizing the Physical Activity of Hospitalized Older People From Wearable Sensors Data Using IoT

Introduction

With the increase in life expectancy, it is now possible to age while remaining active. This is an opportunity, often a grace. However, old age nevertheless leads to physical and functional frailty. Thus, the elderly increased risk of falling. Falls of the elderly represent a major public health problem, both because of the seriousness of their often dramatic medical consequences and because of their undeniable social and economic impact. Therefore, Everything possible must be done to enable these people to continue their lives in the best possible conditions. For this, the idea of preventing falls is necessary to reduce their number and their undesirable consequences. challenges society and technology to find better ways to mitigate the occurrence of such costly and detrimental events as falls.

Today’s Internet of Things (IoT) domain is increasing rapidly (Abdel-Basset, 2019), It is “a global infrastructure for the data society, enabling advanced services by interconnecting (physical and virtual) things based on existing advancing interoperable data and communication technologies” (Rec, 2012), IoT is one of the major technological developments of our times given its potential is fully realized (Farahani, 2018). A large amount of data are emerging every day to be part of the IoT infrastructure. According to Machina Research (Global, 2019), 27 billion connected devices are expected by 2024, while according to Cisco’s report (Cisco, 2017), there will belong nearly 1.5 mobile devices per capita by 2020, and more than 601 million wearable devices will be in use. Connecting all the objects (Jing, 2014) and forming a network of devices is the basic idea of IoT. A major objective of IoT is to make the environment around us smarter, by giving the environment the information it needs, the IoT uses the internet to connected devices that can be easily monitored and controlled also the same things can be automatically detected by other things, further communicate with each other through the internet, and can even make decisions themselves (Tsai, 2014). Over time, various sensory data are collected and generated by an enormous amount of sensing devices. This Will result to generate in a big amount of data from the sensors used for collecting the data. To prevail over these applications some meaningful information must be deduced out of the collected data to make decisions. Applying analytics over such data streams to make control decisions, discover new information and foresee future insights is a pivotal procedure that makes IoT a worthy paradigm for businesses and a quality-of-life improving technology. Among the most extremely useful technologies is Data Mining. A major challenge in these settings is the timely analysis of large amounts of data (big data) to produce decisions and highly reliable and accurate insights so that IoT could satisfy its guarantee.

The aim of this work is the use of Artificial Intelligent to combat falls risks of older folks and to enable these people to continue their lives in the best possible conditions. In this paper, we present a methodology based on multimodal sensors to configure a simple, comfortable and fast fall detection and human activity recognition, and in cases of concern, alerts are sent to caregivers or family members to enable appropriate interventions. a system that can be easily implemented and adopted. In this work, we check whether the traditional data mining algorithms would likewise work for the IoT datasets, or new families of data mining are required. To this end, in this study, we examine the applicability of three data mining algorithms for real IoT datasets. These include K-nn, Naive Bayes, Decision tree The main contribution of this work is the analysis of the efficiency and effectiveness of three of the data mining. We began our work with some related works done in this field, after that in the third section we detailed a description of our system, which will be followed by a presentation of the experiment, these algorithms are analyzed and a comparative study is undertaken to find the classifier that performs the best analysis on the dataset obtained, using a set of predefined performance metrics to compare the results of each classifier and finally conclusions are given.