Footstep Detection using Dempster-Shafer Theory

Pedestrian localization remains a hot topic in confined environments, where satellite signals are often weakened by obstacles and radio frequency technologies are limited due to distance and coverage area constraints. An interesting alternative is to use a localization system based on inertial sensor technology, which can locate a pedestrian by detecting steps and estimating their length and walking angle. This paper proposes a new approach to step detection by combining inertial sensor data using the Dempster-Shafer theory. This approach makes it possible to represent and consider the imperfections in inertial data from accelerometers and gyroscopes and fuse them for decision-making on the validity of the detected step. The results obtained from the experimental evaluation show that the proposed approach performs well in terms of the number of detected steps and outperforms some methods proposed in the literature for step detection.