Robust plane-wave decomposition of spherical microphone array recordings for binaural sound reproduction

Rendering binaural signals from spherical microphone recordings is becoming an increasingly popular approach, with applications in telecommunications, virtual acoustics, hearing science, and entertainment. Such binaural signals can be generated from a plane-wave decomposition of a sound field measured by a spherical microphone array. This process may exhibit ill-conditioned transformations when performed at low frequencies and using high spherical-harmonics orders, thus resulting in a poor robustness to measurement inaccuracies and noise. Previous studies have addressed this issue by employing standard regularization techniques, such as diagonal loading and radial filter limiting. In this paper, we propose an optimal frequency-dependent regularization method that balances system robustness to measurement noise against accuracy of plane-wave decomposition. Unlike previously suggested approaches, the proposed method analytically relates the measured signal-to-noise ratio to the corresponding regularization parameters, hence facilitating a means for controlling the regularization process using a closed-form expression. The method is compared to previously suggested regularization techniques in terms of spatial, temporal, and spectral effects on the resulting binaural signals. Objective results, which illustrate the improved performance of the proposed method, are complemented with a subjective validation of the regularized signals.