Binaural synthesis using a spherical microphone array based on the solution to an inverse problem

Binaural technologies can convey rich spatial auditory information to listeners, using simple equipment such as headphones. Head-related transfer function (HRTF) datasets and spherical microphone arrays are important components to realize advanced binaural recording and reproduction methods. One of these methods is binaural Ambisonics, which captures and reproduces a three-dimensional sound space based on a spherical harmonic analysis. We have proposed an advanced binaural reconstruction method without relying on an intermediate sound field representation. In this method, known as SENZI (Symmetrical object with ENchased ZIllion microphones), individualized binaural signals are directly generated from the microphone recordings by applying weighting filters, which are calculated by inverting a linear system. The accuracy of the sound space synthesized by SENZI is higher at the head-shadow region than that synthesized by conventional binaural Ambisonics. However, the number of microphones imposes a limit on the accuracy at higher frequencies according to sampling theory. This limit can be overcome by making assumptions on the recorded contents or using perceptual information. Because the sensitivity to the phase of human spatial hearing decreases at higher frequencies, we propose utilizing this perceptual knowledge to reconstruct high-frequency contents. Simulation results revealed that high-frequency amplitudes are accurately reproduced using the proposed approach.