Abstract: In the process of navigation and location of a pedestrian for a wearable IMU, the inertial device generates an accumulated drift error affecting the navigation and location accuracy of pedestrian navigation. A multi-stage filtering method is studied: after the zero-speed detection and the zero-speed correction based on Extended Kalman Filter are carried out, the vector domain is divided by the indoor geometric layout features, and the projected matching model is used to determine the optimal coordinates of the nodes to get the trajectory. Using the self-developed MIMU pedestrian navigation module, a field experiment was conducted. The experimental results show that this method can suppress the accumulation of inertial navigation error, which is better than the current single-stage Kalman filter inertial navigation method. There is no pedestrian trajectory passing through the wall.The navigation and location accuracy of pedestrian navigation is improved and the location accuracy is 0.9%. The research has theoretical and practical significance.

Key words: zero-speed detection, EKF, map matching, multi-stage filtering, pedestrian navigation