Utilizing Context Information for Ubiquitous Computation

Abstract

Activity recognition (AR) has gained significant traction in ubiquitous computing in recent years. While AR can be accomplished with a reasonable accuracy in laboratory environment, daily life scenarios pose challenges due to the diversity of the profile of activities and differences in how individuals inherently perform activities. In this work, we propose to define the notion of "context", demonstrate its utility to improve the accuracy of AR, and exhibit its performance in the real-world applications.