Equalization Strategies for Binaural Room Impulse Response Rendering using Spherical Arrays

Reconstruction of binaural room impulse responses (BRIRs) from spherical microphone array measurements in a room and a given head-related transfer function set, is beneficial for binaural reproduction with listener individualization, and for applying head rotations without needing to make numerous measurements of the sound field. Such algorithms often entail the truncation of BRIRs at low spherical harmonic orders, producing reduced reconstruction accuracy at high frequencies and accordingly affecting the perceived timbre. With the aim of restoring high-frequency energy and spatial cues, recent studies have proposed the use of an efficient equalization filter based on a spherical head model, or by direct numerical integration in the space domain. This paper extends the equalization approach by introducing a filter that takes into account the full geometric features of the human head. The proposed filter, as well as previously suggested algorithms, are reviewed and then subjectively evaluated in both measured and simulated environments. A MUSHRA listening experiment is employed to investigate the perceptual effects of equalizing order-truncated BRIRs. Results indicate a clear improvement in timbre, and a positive impact on spatial localization in low order cases.