Combining multiple sparse measurements with reference data regularization to create spherical directivities applied to voice data

Abstract

Obtaining high-resolution, spherical phoneme-dependent directivities for voice radiation is beneficial for numerous acoustics applications. This work reports on a spherical interpolation method based on two interleaved measurements using a regularized least-squares fit with reference data. Maximizing the spherical correlation between previously reported results and measured data determines the regularization hyper-parameters to ensure physical solutions. While the resultant spherical directivity patterns show similarities to time-averaged results, distinct radiation characteristics appear, particularly in the range of 630 Hz to 2 kHz.