Approaches and algorithms for activity recognition have recently made substantial progress
due to advancements in pervasive and mobile computing, smart environments and ambient
assisted living. Nevertheless, it is still difficult to achieve real-time continuous activity
recognition as sensor data segmentation remains a challenge. This paper presents a novel
approach to real-time sensor data segmentation for continuous activity recognition. Central
to the approach is a dynamic segmentation model, based on the notion of varied time …