Fusion of multiple microphone array data for localizing sound sources in an industrial area

Locating sound sources that contribute to noise annoyance near large industrial areas under different meteorological conditions is a hard problem. Permanently installed microphone arrays at the edges of an industrial area allow to determine the direction of arrival of the sound at their location. Several algorithms have been proposed for this purpose yet not all of them are robust against changes in effective sound speed and loss of coherence. Therefore algorithm parameters have to be chosen carefully. In addition, in this paper, a probabilistic approach is proposed to combine the information obtained from three or more arrays. The methodology accounts for the effect of wind and temperature on the direction of arrival. It also estimates uncertainty caused by uncertainty in the local meteorological situation, ground impedance, and presence of typical harbor objects such as stacked containers and piles of coal, etc. The proposed methodology is applied to an industrial area of over 10 kilometer squared and the consistency of detected sources under varying weather conditions is investigated. This shows the validity of the approach.