Suitability Analysis of Wrist-Worn Sensors for Implementing Pedestrian Dead Reckoning Systems

Wearable devices seem an ideal option for implementing pedestrian dead-reckoning (PDR) systems and, eventually, providing seamless location-based services. Among the different available wearable devices, wrist-worn devices, such as smartwatches, have great potential due to their featuring more and more sensors and their convenience for the users. However, the low number of wrist-worn PDR systems proposed in the scientific literature reveals the intrinsic difficulty of estimating the position from the upper limbs. Before starting to design new algorithms for improving wrist-worn PDRs, it was considered necessary to make a deep analysis of the characteristics of the different signals generated in a wrist-worn sensor when a person walks, and identify how they affect the implementation of an inertial PDR system for that body location. After this analysis task, it seems feasible to implement wrist-worn PDR systems thanks to the presence of clear step/stride patterns in the signals coming from different types of wrist-worn sensors. However, these patterns show a non-negligible variability dependent on the different motion modes, walking speed and the way each user moves their upper limbs while walking, indicating that adaptive algorithms and prior calibration stages are necessary.