Multi-attitude-based target localization method

In recent years, location-based services have gradually penetrated all aspects of people's daily life, bringing great convenience to people's work and life. Pedestrian Dead Reckoning (PDR) is a positioning method that can derive a pedestrian's movement trajectory by relying on sensors information, with a high degree of autonomy and continuity. However, existing PDR tends to restrict the target posture to remain steadily pointing in the direction of pedestrian movement throughout the localization process. In addition, although it can achieve a high degree of positioning accuracy in a short period of time, it generates a cumulative error over time. To these problems, this paper proposes a fusion localization method that combines PDR based on multiple smartphones pose recognition and acoustic localization. In this method, we first employ a GWO-ELM based target posture recognition model. Then based on the posture recognition result for PDR localization, Time Difference of Arrival (TDOA) technique is used for acoustic localization. Finally, particle filtering is utilized to combine the two methods. To demonstrate the validity of the method, we performed several experiments on a motion trajectory with a length of 97.2 meters. The results show that the method has better localization performance than the traditional method. In addition, it can realize high-precision localization under different devices with universality and stability.