Predicting localization accuracy for stereophonic downmixes in Wave Field Synthesis

Wave Field Synthesis enables the creation of a correct spatial impression for an extended listening area. However, the synthesized sound scenes are most often created with a model-based rendering approach. This enables, beside other advantages, interactivity of the listener with the sound scene. On the downside it requires anechoic recordings of every sound event of the sound scene, a requirement that most often does not hold for existing music recordings and productions. In this study we investigate methods of reproducing two-channel stereophonic recordings with Wave Field Synthesis. Thereby di erent virtual stereophonic loudspeaker layouts, like point sources and plane waves are arranged. It is further investigated how these interact with di erent geometries of the underlying loudspeaker array for Wave Field Synthesis. As typical setups linear, circular and box shaped geometry is employed. A common drawback of stereophonic reproduction is the sweet-spot for localization. Outside of the sweet-spot the listeners start to localize the reproduced sources towards the single loudspeakers. Recent advantages in predicting the localization for Wave Field Synthesis setups with a binaural model will be utilized in this study. Applying the binaural model the in uences of the downmixing method and the underlying geometry are investigated. Especially the number of active loudspeakers and the usage of plane waves for stereophonic presentation di er from a typical stereophonic setup and potentially allow to increase the sweet-spot size.