Computing interaural differences through finite element modeling of idealized human heads

Abstract

Acoustical interaural differences were computed for a succession of idealized shapes approximating the human head-related anatomy: sphere, ellipsoid, and ellipsoid with neck and torso. Calculations were done as a function of frequency (100–2500 Hz) and for source azimuths from 10 to 90 degrees using finite element models. The computations were compared to free-field measurements made with a manikin. Compared to a spherical head, the ellipsoid produced greater large-scale variation with frequency in both interaural time differences and interaural level differences, resulting in better agreement with the measurements. Adding a torso, represented either as a large plate or as a rectangular box below the neck, further improved the agreement by adding smaller-scale frequency variation. The comparisons permitted conjectures about the relationship between details of interaural differences and gross features of the human anatomy, such as the height of the head, and length of the neck.

On the localization of high-frequency, sinusoidally amplitude-modulated tones in free field

“…The bright spot at the far ear causes a minimum in the ILD function. Therefore, the ILD shows a peak as a function of azimuth.The effects of the bright spot can be calculated analytically for a spherical head, with results that compare reasonably with the more realistic shapes used by Cai et al (2015). Table III shows θ peak , the azimuth of a source for which the peak occurs, measured with respect to the forward direction of the head.…”

HRTF Selection by Anthropometric Regression for Improving Horizontal Localization Accuracy

“…By contrast, the inclusion of shoulder circumference as significant predictor could be seen as counterintuitive, and might be due to moderate correlations with head width (r = 0.68 in the HUTUBS database) and depth (r = 0.38). However, recent work [15] found that the inclusion of the torso in an idealized head model is responsible for the addition of small ripples to the ITD that increase its frequencydependent structure and improve agreement with ground-truth acoustical data especially at middle lateral azimuths.…”

“…Different analytical solutions for individual ITD estimation and adaptation, such as spherical and ellipsoidal head models, are available in previous literature (for a review see [14]). Despite their usability, these models suffer from evident discrepancies between modeled and human interaural differences [15]. Subjective ITD selection procedures [16] are available as well.…”

Deep neural network models of sound localization reveal how perception is adapted to real-world environments

“…For instance, it was supposed for many decades that interaural cues neatly segregated in their function, with time and level differences subserving localization for low-and highfrequency sounds, respectively ("duplex" theory) (14). However, it has become clear that level differences are perceptually important even at low frequencies (35)(36)(37). Here and in other domains of sensory processing, models that can replicate real-world behavior, and thus help reveal the information needed to mediate such behavior, could have a transformative effect on perceptual science.…”