Kalman Filtering for Tracking a Moving Acoustic Source based on Energy Measurements

The present work addresses the problem of continuous localization of a moving acoustic source based on energy measurements. This type of measurements is of particular practical interest since it requires less hardware than the ones based time and/or angle. The work intends to demonstrate that, even though the measurement model is highly non-linear, it can be maneuvered towards a framework favorable for applying Kalman filtering (KF) techniques. As such, we propose two approaches based on a Kalman filter and an Extended Kalman filter (EKF), that show superior tracking performance in comparison with sequential maximum likelihood estimator (MLE). After analyzing the computational complexities of the proposed methods, a set of simulation results is presented showing that, for different trajectories and different sets of sensors, the KF approach excels the MLE (which neglects any prior state knowledge information and makes use of observations only) with a decrease of the localization error from 34% to 40%, while the EKF achieves a reduction from 65% to 77%.