Abstract
A far-field random array is implemented for localization and separation of noise sources. Microphone positions are optimized, with the aid of the simulated annealing (SA) method as a supervised Monte Carlo approach, random samples of sensor position are drawn from Gaussian distribution to minimize the side-lobe maximum of the far-field beam-pattern. A two-stage algorithm is devised on the basis of the spherical-wave model on the image plane. In the localization stage, the active source regions are located by using the delay-and-sum (DAS) method, followed by parametric array localization with improved resolution. In the separation stage, source amplitude extraction can be achieved by formulating an inverse problem based on the steering matrix relating the sound pressures received by the microphones and the source amplitudes. The number of sources is selected to be less than the number of microphones to render an overdetermined problem which can be solved by using the Tikhonov regularization (TIKR) and Maximum Likelihood Estimation (MLE). Alternatively, the separation problem can be augmented into an underdetermined problem which can be solved using the compressive sensing (CS) method. Numerical and experimental results are presented to validate the proposed technique.