A bio-inspired robotic sound localization method

Abstract

Audio is an important means to improve interactive performance between robot and environments. Aiming at developing effective means to guide robotic navigation in domestic environments, this paper presents a biologically inspired method for robotic sound localization. In contrast to the popular use of time-to-arrival differences from microphone arrays, the proposed approach is based on binaural cues selection from only a pair of microphones recording similar to the principles observed in mammalian hearing system. Especially, with a newly developed robotic head model, the localization algorithm is able to estimate sound source directions more accurately and reliably beyond the limits of the robotic auditory system. The preliminary experimental results show that higher performance can be achieved using the proposed method.