Optimal positioning of a binaural sensor on a humanoid head for sound source localization

摘要

A generic approach to the placement of a binaural sensor on a humanoid robot head is proposed in order to improve sound localization. After a brief description of binaural and spectral cues, the method is first illustrated on the well-known head spherical approximation, which response can be described analytically. Then, a simplified CAD model of a mannequin is considered. It is argued that recent developments on analytical expansions of HRTFs can help to get a solution even from a limited set of simulated head responses. The obtained conclusions noticeably comply with some antropomorphic statistics.