Algorithm based on particle filter framework for personal moving state classification

Abstract

Nowadays, positioning data can be obtained using various mobile devices. These data can be used to determine speed, along with latitude, longitude, and a timestamp. However, because of the positioning system errors, the speed value itself cannot be used to categorically determine whether the corresponding device is in motion or if it is stationary. To identify the mobility state of the device, an algorithm based on a particle filter is introduced in this paper. The particle filter is used to filter the speed values, and the mobility state is identified probabilistically. A framework similar to sequential importance sampling based particle filter algorithm is utilized, and a parameter determination process for human mobile speed distribution is introduced as an indirect derivation method of weight for particle filter. Based on the sliding windows of speed values, time weighted speeds and parameters for human mobile speed distribution are determined in real time. The algorithm and related experimental results are also presented. The execution time for this algorithm is sufficiently small so as to be applicable for real time application, even for the positioning dataset for one day. Three different methods are used to calculate the weight of the particle filter. A noteworthy characteristic of this algorithm is that it only requires a positioning data set and does not require any other data or equipment.