Estimation of room dimensions from a single impulse response

Abstract

In this paper we propose a methodology for the estimation of the geometry of an environment based on a single Acoustic Impulse Response (AIR). The estimation algorithm makes use of tools for the modeling of propagation based on geometrical acoustics. A suitable cost function evaluates the distance between the simulated and measured AIRs. The room minimizing the cost function is chosen as the correct one. The cost function is strongly non linear. As a consequence, in order to reduce the complexity of the minimization problem, the algorithm needs a hypothesis about the class of geometry of the environment under analysis, such as rectangular or L-shaped rooms. We prove the effectiveness of the proposed algorithm with a number of simulations with increasing complexity.