Affiliation of Author(s): 北京航空航天大学

Journal: IEEE Transactions on Instrumentation and Measurement

Key Words: Frequency modulation, machine learning (ML),pedestrian dead reckoning (PDR), regression model, step length estimation (SLE)

Abstract: Among various indoor positioning methods, pedestrian dead reckoning (PDR) has become one of the mainstream methods for requiring neither expensive infrastructure nor laborious surveys. Step length estimation (SLE) is one of the key components of PDR. Most of the existing SLE methods utilize acceleration or angular velocity to estimate the step length, which are susceptible to the measurement noise of low cost inertial sensors as well as different walking speeds and persons. In this article, we propose an SLE method based on the adaptive combination of frequency-modulated (FM) radio signal and acceleration, which considers both accuracy and practicality.First, based on the propagation model theory of radio signals,we derive the relationship between the received signal strengthindicator (RSSI) of FM signal and step length, which provides the theoretical basis for the proposed SLE algorithm. Second, FM signal features related to step length are extracted. Third, since the availability of FM signal depends on the type of scenes, an availability assessment strategy is proposed, enabling the utility in various scenes. Finally, with the assistance of availability assessment results, we apply support vector regression (SVR) to realize SLE by combining the FM signal features and acceleration features adaptively. Field experiments conducted by 14 experimenters in different indoor scenes validate that the introduction of FM signal features effectively reduces the SLE errors, thus improving the positioning performance.

Indexed by: Journal paper

First-Level Discipline: Information and Communication Engineering

Document Type: J

Translation or Not: no

Included Journals: SCI