Subtractive Clustering as ZUPT Detector

Inertial-based indoor pedestrian tracking that uses Micro electro mechanical Systems (MEMS) technology suffers undesirable positional drift over time. As widely attested, zero-velocity updates (ZUPT) from the stance phase reduce the error growth from a third order polynomial to a linear one. However, researchers are struggling to find consistent ZUPT, especially when the pedestrian walks naturally, which has changes in walking speed or unpredictable pauses. In this paper, a novel approach to extract the ZUPT based on subtractive clustering is proposed and discussed. Its performance is compared to other techniques using internally collected and publicly available datasets. The results show that the proposed method outweighs the others in providing consistent performance level.