Robust 3D Robotic Sound Localization Using State-Space HRTF Inversion

Abstract

We address the problem of robotic real-time binaural hearing using a humanoid head. The sound detection ability for a previously proposed robotic 3D binaural sound localization technique is stabilized and considerably enhanced. The proposed technique is based on using two small microphones placed inside the ear canal of a robot dummy head mounted on a torso. As initially proposed, the detection algorithm is based on a simple correlation approach using a generic set of head related transfer functions, HRTFs, inverted using the fast Fourier transform, FFT, method. In the current work, we replace the Fourier inversion mechanism with state-space inversion using outer-inner factorization, and we demonstrate the performance through simulation and further in a household environment. Due to this modification, our robotic sound detection set up proves to be more noise-tolerant and able to localize sound sources in a three-dimensional space with a higher precision.