How round is the human head?

Abstract

Binaural microphone arrays are becoming more popular for hearing aids due to their potential to improve speech understanding in noise for hearing impaired listeners. However, such algorithms are often developed using three-dimensional head-related transfer function measurements which are expensive and often limited to a manikin head such as KEMAR. As a result, it is highly desired to use a parametric model for binaural microphone array design on a human head. Human heads have been often modeled using a rigid sphere when diffraction of sound needs to be considered. Although the spherical model may be a reasonable model for first order binaural microphone arrays, recent study has shown that it may not be accurate enough for designing high order binaural microphone arrays for hearing aids on a KEMAR (Merks et al., 2014). In this study, main sources of these errors are further investigated based on numerical simulations as well as three-dimensional measurement data on KEMAR. The implications for further improvement will be discussed.