DOA Acoustic Source Localization in Mobile Robot Sensor Networks

This paper explores the Audio Source Localization problem (ASL) using only the Direction of Arrivals (DOA) estimated by a Mobile Robot Sensor Network of a fixed single acoustic source. It also proposes a fast algorithm for localizing the 2D source position using the recent Gaussian Probability over DOA approach (GP-DOA). More specifically, the paper focuses on the analysis and explanation of the proposed algorithm that runs in Theta(log(2)(2) n) time instead of the state-of-art algorithm that runs in Theta(n(2)). Several simulation tests varying the errors over the DOA estimation and over the position of the robots are done. Test results show that the proposed approach can reach the same precision of the GP-DOA approach in less time and outperforms the Weighted Least Square method (WLS-DOA) if the maximum error over the positions of the robots is under 0.4 m in a 100 m(2) room. A real experiment using Microsoft Kinect as DOA-sensors and a Pioneer 3-AT robot within the ROS framework shows that the algorithm can be a powerful approach in robotics for ASL.