B-format for binaural listening of higher order Ambisonics

Abstract

B-format is a four-channel signal capable of rendering a sound scene with spatial information. It can be regarded equivalent to first-order Ambisonics. Ambisonics requires a high order to contain precise spatial information, and higher-order Ambisonics requires an exponentially large amount of data. This limitation comes from the fact that the original aim of Ambisonics is to reproduce the whole sound field. However, as mobile devices are prevalent, users often listen to sound media through earphones. Because nowadays users can hold sound contents individually, one can assume that sound contents could be produced adaptively to each user. Here we propose a method to make B-format signals more suitable for individual binaural listening. We assume that the production side can capture a sound scene with higher-order Ambisonics, because it may be processed for enterprise applications. Under this assumption, the binaural signal is once generated from the higher-order Ambisonics, and then its B-format signal is obtained by inversely processing the signal, assuming the first-order Ambisonics. Computer simulations show that interaural phase differences (IPDs) are improved at a frequency region where IPD dominantly affects sound localization. Results of hearing tests are also discussed.