2D sound source mapping from mobile robot using beamforming and particle filtering

This paper describes a particle filter based sound source mapping system that builds 2D sound source maps from directional sound readings taken from a mobile robot. The method uses a sound source localization model that is represented by Gaussian distribution for both direction and distance. To do this, accurate directional localization of sound sources is required, and two key components have been developed to achieve this: 1) a 32ch low side-lobe microphone array that is designed by beam forming simulation to have a) an omni-directional response, b) a narrower main-lobe, and c) lower side-lobes, in 700-2500[Hz] acoustic signals; 2) directional localization of different pressure sound sources by combining the Delay and Sum Beam Forming (DSBF) and the Frequency Band Selection (FBS) methods. Finally, experimental results show the proposed method can map sound sources in two dimesionals with high accuracy (less than 50[cm] error).