Robust source localization using decision-directed algorithm and confidence weights in Unattended Ground Sensors system

Abstract

Seismic Unattended Ground Sensors (UGS) systems have a major role in the developing area of seismic signal processing, with applications mainly in security and surveillance systems. Identifying and localizing a potential threat is a preliminary requirement in such systems. Array processing based on measured time of arrivals or gain-ratio values is widely used for solving the localization problem. However, for real world seismic data, estimating time differences and gain-ratios of arrival is a difficult task, due to both the nature of sensors networks and of seismic signals. Sensors synchronization is a common difficulty in networks and the demand for low power consumption and transmission rates prevents solving it by cross-correlating the signals. High variations in sound velocity and background noise among different types of ground, which characterize the underground environment, are additional factors for these difficulties. Hence, applying direct localization algorithms on seismic data often proves ineffective. In this paper, a novel approach toward seismic source localization using UGS system is presented. Given an event of recurring nature, the proposed algorithm is based on two principles which increase its robustness. First, it utilizes both time differences and gain-ratios measurements in a decision directed process. In addition, confidence weights are assigned for each recurrence of the event thus further performance improvement is achieved. Results for applying the proposed algorithm on real-world seismic data are presented and the advantages of the proposed algorithm are demonstrated.