Human behavior sensing: challenges and approaches

Abstract

In recent years, Activities of Daily Living Scale (ADLs) is widely used to evaluate living abilities of the patients and the elderly. So, the study of behavior sensing has attracted more and more attention of researchers. Behavior sensing technology is of strong theoretical and practical value in the fields of smart home and virtual reality. Most of the currently proposed approaches for tracking indicators of ADLs are human-centric, which classify activities using physical information of the observed persons. Considering the privacy concerns of the human-centric approaches (e.g. images of home environment, private behavior), researchers have also proposed some thing-centric approaches, which use environmental information on things (e.g. the vibration of things) to infer human activity. In this paper, by considering the unified steps in both the human-centric approaches and the thing-centric approaches, we make a comprehensive survey on the challenges and proposed methods to do behavior sensing, which are signal collection, preprocessing, feature extraction, and activity recognition. Moreover, based on the latest research progress, we post a perspective from our standpoint, discussing future outlook and challenges of human behavior sensing.