Robust speaker localization system for a humanoid robot in noise and reverberation environments

Abstract

Human-robot interaction (HRI) is an important study to allow more natural and effective communication between humans and robots. Recently, research on HRI has been receiving an increasing amount of attention. In this paper, we report on a speaker localization system for a humanoid robot, as part of an auditory-based system of HRI. Our system uses a six-channel microphone array on a humanoid robot's head. To achieve good results in environments of noise and reverberation, we utilize a pre-emphasis filter; a voice activity detection (VAD) algorithm; and generalized cross correlation with phase transform weighting (GCCPHAT) method. Experimental results show that our speaker localization system for a humanoid robot can successfully localize a speaker's position over the entire azimuth and altitude in noise and reverberation environments.