Resident activity recognition based on binary infrared sensors and soft computing

The basic concept of a smart space (SS) is to be aware of the context information related to environmental and human behavioral changes, and to provide appropriate services accordingly. To obtain context information, we may use video cameras, microphones, and other monitoring devices. Although these devices can obtain complex environmental data, they are not suitable for building private smart space (PSS) because of the privacy issue. Human users do not like being monitored in their private spaces. In this study, we investigate the possibility of recognizing certain activities using binary data collected by using infrared sensors. Infrared sensors have been used mainly for detecting the existence/absence of the residents in a region of interest. Here, we consider four types of activities, namely, No-Activity, Very-Weak-Activity, Weak-Activity, and Strong-Activity. Our main goal is to provide a way for building PSS using low-cost and non-privacy-sensitive devices. We have conducted some primary experiments by collecting user activity information using binary infrared sensors. Generally speaking, activity related sensor data are sensitive to various factors. To effectively address this issue, we propose a recognition method based on fuzzy decision tree. The results of the primary experiments show that the recognition rate of proposed method can be as high as 85.49%. The results are encouraging, and show the possibility of building PSS using binary infrared sensors.