Estimation of reflective surfaces from continuous signals

The geometry of an enclosure is of interest for example when synthesizing impulse responses or when studying concert halls. Recently, several approaches for estimation of the room geometry, or the reflective surfaces, have been proposed. These approaches use a priori information of the source signal for the room geometry estimation, typically, impulse response measurements. Here, a method for estimating the reflective surfaces from continuous signals, such as speech or music, is proposed. The method is based on inverse mapping of the acoustic multi-path propagation problem. The validity of the method is demonstrated in a real auditorium. With reasonable signal-to-noise ratio the proposed algorithm has less than 0.05 m error in the position of the point of reflection and less than 1 degree of error in the direction of the normal of the surface.