Improved pedestrian tracking through Kalman covariance error selective reset

Kalman filtering is one of the most widely used approaches to handling inertial sensors in pedestrian tracking systems. This technique uses a covariance error matrix to estimate position. This reported study leads to the hypothesis that there is no correlation between some elements of this matrix from one step to the next. Therefore, a selective reset of these elements at the end of each step improves position estimation. A set of these elements is proposed, and a statistical study is conducted using 32 data traces from the same path. Four parameters are analysed: the correction mean length, the position error, the altitude error and the travelled distance. As a result, all of these parameters obtain a loose statistical significance when the covariance error selective reset is applied.