The Gaussian Processes for Acoustic Localisation and Tracking in Wireless Sensor Network
R. Azzam, Nabil Aouf · 2013
In this paper, the problem of enhancing the robustness of acoustic source localisation and tracking based on Time Delay of Arrival (TDOA) in wireless sensors networks (WSN) is considered. Distributed acoustic localisation is used relying on the sensing and the communication specification of the motes. Parameter measurements such as the detection time that is provided by the distributed sensors can be erroneous and lead to inaccurate target locations. Gaussian processes (GPs) are flexible and fully probabilistic methods that we propose to adapt in conjunction of Unscented Kalman Filter (UKF) to improve target tracking performances for WSN. Indeed, GPs with their capacity of dealing with prediction of quantities under uncertainty revealed to contribute greatly into providing good results for the problem exposed in this research.