Pedestrian Dead Reckoning Based on Motion Mode Recognition Using a Smartphone.

Area of Science:

• * Computer Science
• * Electrical Engineering
• * Navigation Systems

Background:

• * Pedestrian dead reckoning (PDR) is crucial for indoor and outdoor localization.
• * Traditional PDR methods using smartphones face challenges with motion mode recognition and heading accuracy.
• * Principal Component Analysis (PCA)-based methods can be influenced by non-horizontal accelerations, affecting direction vector accuracy.

Purpose of the Study:

• * To develop an accurate PDR approach using smartphone motion mode recognition.
• * To propose an improved PCA-based method (PCA-GA) for robust pedestrian heading inference.
• * To enhance heading accuracy by addressing ambiguities related to phone pose.

Main Methods:

• * Implemented motion mode recognition using Support Vector Machine (SVM) and Decision Trees (DT) to classify pedestrian movement states and phone poses.
• * Developed a PCA-based method with global accelerations (PCA-GA) to accurately infer pedestrian headings.
• * Incorporated an ambiguity elimination method based on phone pose analysis for heading calibration.

Main Results:

• * Achieved a 92.4% recognition accuracy for combined movement states and phone poses.
• * Demonstrated high heading estimation accuracy with 50% and 75% absolute errors of 5.6° and 9.2°, respectively.
• * Reduced localization error to approximately 3.5 meters over a 164-meter trajectory.

Conclusions:

• * The proposed PCA-GA method significantly outperforms traditional PCA-based and heading offset methods in accuracy.
• * Smartphone-based motion mode recognition combined with PCA-GA offers a reliable solution for accurate PDR.
• * This approach effectively mitigates localization errors across diverse movement patterns and phone orientations.