HUMAN ACTIVITY MONITORING USING MACHINE LEARNING for IoT BASED SECURITY and HEALTHCARE APPLICATIONS

The internet of things (IoT) has formed a new avenue of research due to the advancement of information and communications technologies and has redefined the entire research area of Human Activity Recognition (HAR). One of the Wi-Fi signal properties, Channel State Infor-mation (CSI), can be used to identify different human activities. To meet future needs, activity recognition is in demand, whether it is for security applications, smart home facilities and elderly care monitoring. In this paper, a prediction model for human activity recognition is developed using machine learning algorithms. The proposed system works at 3.75 GHz for single subject activity recognition using the CSI signatures. First, the acquired CSI data associated with each recorded activity instance are processed. Then the relevant features are extracted, and the es-sential features are selected. Finally, the selected features trained in different machine learning algorithms to identify the different performed activities. After an extensive, methodical search, we developed a system that can predict daily activities (sitting, standing, and walking) with an overall accuracy of 93% for single-subject activity monitoring in security, healthcare, and other applications. We believe our model can accurately predict the subject's activities and can therefore assist the users in predicting and making emergency decisions, especially in a limited resource scenario, based on the activity of the subject and their future potential risk of losses.