Foot-Mounted Pedestrian Navigation Method Based on Gait Classification for Three-Dimensional Positioning

Abstract

A foot-mounted pedestrian navigation system (PNS) that uses MEMS inertial measurement units (IMUs) can track the position of a person when walking normally on two-dimensional (2D) plane. However, the positioning accuracy is decreased under complex gaits. This article proposes a pedestrian navigation method based on gait classification for three-dimensional (3D) positioning. The commonly used gaits are divided into seven types and a bidirectional long short-term memory recurrent neural network (BLSTM-RNN) is constructed as gait classifier. When the detected gaits are walking sideways or running, the step length models are built to calculate single step length. Otherwise the inertial navigation algorithm with zero velocity update (ZUPT) is used to calculate travel distance. If the pedestrian is going upstairs or downstairs, a barometer together with stair height constraint is applied to obtain altitude variation. Meanwhile, the heading drift is reduced combining zero integrated heading rate (ZIHR) method and simplified heuristic drift reduction (HDR) method. The gait classification results show that the classification accuracy of designed BLSTM-RNN reaches 99.4%. The results of mixed gaits experiment and multi-floor experiment show that the positioning errors of 2D and 3D trajectories are less than 2%.