期刊名称：International Journal of Multimedia and Ubiquitous Engineering
印刷版ISSN：1975-0080
出版年度：2015
卷号：10
期号：11
页码：69-84
DOI：10.14257/ijmue.2015.10.11.07
出版社：SERSC
摘要：Most of pedestrian inertial navigation system estimates displacement based on the integration of inertial sensors measurements. However, due to low-cost sensors and pedestrian dead reckoning inherent characteristics these systems provide huge location estimation errors. To suppress some of these limitations we propose a pedestrian inertial navigation system based on low-cost sensors and on information fusion and learning techniques. The proposed system introduces a step characterization module that characterizes the step according to the activity that the pedestrian is performing. This module performs three characterizations: terrain, direction and length. Thus, in this work are presented and evaluated several machine learning approaches that perform the terrain characterization. The inclusion of this machine learning module led to a significantly better performance of the pedestrian inertial navigation system.
