Activity Recognition Using Ubiquitous Sensors

Abstract

With the unprecedented sensing capabilities and the emergence of Internet of things, studies on activity recognition have been hot issues for different application areas, such as pervasive healthcare, industry and commerce, and recommendation systems. Much effort has been devoted to activity recognition using different sensors. Based on the differences of ubiquitous sensors, the authors classify the existing work into approximating sensing, wearable sensing, and video/audio sensing. Generally, methodologies for activity recognition are divided into logical reasoning and probabilistic reasoning. They illustrate the generalized framework and outline the advantages and disadvantages for each algorithm. Despite the research on activity recognition, activity recognition still faces many challenges in many aspects including nonintrusive data collection, scalable algorithms, energy consumption, and semantic extraction from social interaction. Towards those challenging research issues, the authors present their contributions to the field of activity recognition.