Abstract
Position estimation of sound sources using a two-dimensional passive array is useful in a limited number of applications such as indoor robotics. A new method is described that intelligently ranks and evaluates multiple microphone configurations within the array in order to decide which pairs of microphones to use in order to produce accurate position estimates. The method segments a two-dimensional array into various combinations and configurations of microphone pairs and flattens these configurations into one-dimension for comparison. Each configuration is ranked based on the microphones' spatial information, which is known a-priori, and the incoming bearing estimates for the source. These rankings are then applied to select different microphone pair configurations, whose position estimates are combined in a weighted average to improve the source position estimate calculation. Simulations and experimental data show that this method can select microphone configurations that provide the least position error, and by combining their individual data, position accuracy of the array can be improved.