Pervasive Activity Logging for Indoor Localization in Smart Homes

Abstract

The scientific contribution of this work to address multiple global societal and economic challenges associated with the increasing aging population is primarily two-fold. First, it presents and discusses a new computing methodology - Pervasive Activity Logging that involves the creation of an adaptive and semantic collection or 'log' of human activities in Internet of Things (IoT)-based pervasive environments, along with the characteristics of user interactions associated with these activities in terms of the contextual, behavioral, spatial, and temporal information. Second, by using this concept of Pervasive Activity Logging, this paper presents a novel machine learning and pattern recognition-based approach for Indoor Localization during different activities performed within the premises of an IoT-based pervasive environment, such as a Smart Home. This learning approach used a Random Forest-based classification model and detected the user's indoor location with an accuracy of 83.02%. These methodologies were developed by integrating the latest advancements from Pervasive Computing, Big Data, Information Retrieval, Internet of Things, Human-Computer Interaction, Machine Learning, and Pattern Recognition. The results presented and discussed uphold the relevance, potential, and importance of these methodologies for contributing towards independent living, healthy aging, and improved quality of life of elderly people in the future of IoT-based pervasive environments, such as Smart Homes.