Magnitude Least-Squares Based Ambisonics Estimation of Head-Worn Device Microphone Measurements for Binaural Reproduction

Immersive audio experiences in virtual environments rely heavily on accurate Higher-Order Ambisonics (HOAs) estimation and its binaural rendering through Head-worn Devices (HwDs). Addressing the irregular geometry and unknown scattering effects of HwDs, HOA estimation using Wearable-Device-Related Transfer Functions (WDRTFs) was introduced. However, the limited microphone measurements forced the truncation of higher-order WDRTFs which are critical at high frequencies. To alleviate its perceptual impact, this paper proposes a Magnitude Least-Squares (MagLS) based preprocessing of WDRTFs for HOA estimation. A MUSHRA-based listening test showed significant improvement in the binaural signal quality with promising results in comparison to a commercially used spherical microphone array.