A robust context-based heading estimation algorithm for pedestrian using a smartphone

摘要

Heading estimation is a crucial part of pedestrian dead reckoning (PDR), for a small bias of heading may lead to a huge localization error. Several methods have been proposed for heading estimation in PDR, but most of them are just available when the smartphone are carried fixedly. In this paper, we proposed a context-based heading estimation algorithm which is suitable in several contexts, such as texting, calling, swinging, or just in a pocket, etc. To achieve this goal, we first proposed an adaptive attitude and heading reference system (A-AHRS) to get the heading and attitude, which is essential for heading estimation. In contexts where the offset angle between the sensor's yaw and pedestrian's heading is fixed, we compute the heading by finding this offset and using a rotation matrix to compensate this offset. In other cases where the yaw angles change dynamically, it's not possible to fix the offset of heading. In this paper, we applied a filter to mitigate the heading noise due to pedestrian's periodic motion patterns in the frequency domain. Then we can get a robust heading from the filtered yaw angles. Results show that the proposed solution outperforms the conventional methods in terms of robustness and accuracy.