A Model with Hierarchical Classifiers for Activity Recognition on Mobile Devices

Activity recognition based on mobile devices has been used in a wide range of applications, including health monitoring, mobile context-aware and inertial positioning. However, the single-layer classifier model cannot accurately recognize users' physical activity due to the diversity of activities. This paper described and evaluated a model with hierarchical classifiers for activity recognition. To implement the hierarchical model, a reasonable and effective pattern combination algorithm based on similarity between activities was put forward to design the structure of hierarchical classifiers. A new concept of confusion ratio was defined to measure the similarity between activities. The experimental results show that the activity recognition model using hierarchical classifiers achieves a good performance.