A Handheld Inertial Pedestrian Navigation System With Accurate Step Modes and Device Poses Recognition

In this paper, a handheld inertial pedestrian navigation system (IPNS) based on low-cost microelectromechanical system sensors is presented. Using the machine learning method of support vector machine, a multiple classifier is developed to recognize human step modes and device poses. The accuracy of the selected classifier is >85%. A novel step detection model is created based on the results of the classifier to eliminate the over-counting and under-counting errors. The accuracy of the presented step detector is >98%. Based on the improvements of the step modes recognition and step detection, the IPNS realized precise tracking using the pedestrian dead reckoning algorithm. The largest location error of the IPNS prototype is ~40 m in an urban area with a 2100-m-long distance.