Indoor sound source localization using a room feature-based method

Indoor sound source localization is one of key foundations for public security, source separation, surveillance, etc. Instead of neglecting the shape and boundary information of the room in the traditional methods, a room feature-based localization method is proposed in this paper. This method uses the geometry information as the prior knowledge to construct a numerical model which simulates the characteristics of the indoor sound field. By dividing the room into a number of subdomains using nodes, the Helmholtz equation can be inversely solved using the numerical model based on the measured signals. This inverse result finally gives the location of the sound source which is represented by the location of the subdomain. The method is validated by numerical simulations. It is shown that the method can achieve high localization accuracy even in a non-convex room with the assistance of the room geometry. The verification also demonstrates that the method is valid for the multiple source localization.