An Interpretable Orientation and Placement Invariant Approach for Smartphone Based Activity Recognition

摘要

Traditional activity recognition methods based on accelerometers generally assume that the orientation and placement of the sensing devices are fixed. However, the performance of smartphone based activity recognition tends to be greatly affected by the uncontrolled orientation and placement. In this paper, a novel activity recognition approach is proposed, which is based on smartphone accelerometers and is independent of device orientation and placement. Specifically, our approach integrates coordination mapping and shape lets techniques, which make the approach to be reliable and interpretable for activity classification. Experimental results show that our approach has its superiority on reducing the negative effects of uncontrolled device orientation and placement, and is able to provide an insight into classification results.