A real-time step detection algorithm based on dynamic peak-valley change threshold screening and dual-time threshold constraints

Abstract

As indoor inertial navigation, Pedestrian Dead Reckoning (PDR) has attracted much attention because its positioning results are not affected by the signal of smartphones. Due to the accumulated error of PDR, it is mostly used to improve indoor positioning accuracy by combining with absolute positioning technologies such as Wi-Fi and geomagnetism. The accuracy and efficiency of step detection in PDR have an important impact on the fusion positioning system. The existing step detection algorithms and the built-in pedometer of Android smartphones have the phenomenon of the delayed output of step detection results, which affects the efficiency of real-time indoor positioning based on PDR. In this paper, a real-time step detection algorithm based on dynamic peak-valley change threshold screening and dual-time threshold constraints is proposed. First, the true peak and true valley accelerations are screened according to the amplitude of peak-valley change based on the dynamic peak drop threshold and valley rise threshold, to eliminate the influence of noise near peak-valley values. Then, the peak-valley accelerations are further determined according to the dual-time threshold constraints. The accuracy and efficiency of step detection are effectively improved. This paper has completed experiments on the self-developed Android APP, and the experimental results show that the accuracy and efficiency of step detection are better than the pedometer provided by smartphones.