Abstract
The availability of inertial navigation sensors in smartphones has facilitated the development of pedestrian dead reckoning (PDR) models on a large scale. These models often consist of a step detection algorithm combined with a heading estimation routine. Common approaches to step detection include searching for peaks/valleys in the acceleration signal, principle frequency estimation, and machine learning techniques. Since the sensors embedded in smart devices are prone to noise, the position error grows unbounded if unchecked and requires periodical corrections based on external measurements. In this work, we propose a novel step detection algorithm based on sine-wave approximation of the acceleration signal. This method detects step fractions as well as full steps, which allows for continuous and real-time updates. The step detection algorithm is combined with a heading estimation routine described in our previous work to obtain a stand-alone PDR model. To mitigate error accumulation, we fuse the proposed model with position and heading measurements provided by a commercial indoor positioning system based on ultrasound. We evaluate the performance of the PDR and fused model in an open office environment, by walking along a trajectory while carrying a smartphone in hand or in the pocket. The results demonstrate the feasibility of the sine-wave approximation approach to step detection, as well as the expected benefits of fusing PDR with the ultrasonic system.
Highlights
S MART devices have boosted the development of Location Based Services (LBS) in recent years, due to their diverse set of embedded sensors and widespread availability
We present a novel step detection algorithm based on sine-wave approximation of the accelerometer signal
The test area was equipped with a Motion Capture (MoCap) system which provided a ground truth reference of the walking trajectory
Summary
S MART devices have boosted the development of Location Based Services (LBS) in recent years, due to their diverse set of embedded sensors and widespread availability. An INS provided a position estimate based on the step count derived from accelerometer data, and heading based on the compass These inertial measurements were inputs to the propagation phase of an Extended Kalman Filter (EKF). In case of no coverage by the ultrasonic system, the inertial sensors from the smartphone provided the position, calculated from the step length and orientation. This ultrasonic system was later used in [30] to fix the position estimates provided by an INS based on a Fuzzy Inference System (FIS). To enable navigation in areas with limited ultrasonic coverage, an external INS was attached to the user’s leg, and connected to the device via Bluetooth It calculated the step length from the pitch signal, obtained from an EKF.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call