Loudspeaker Directivity and Playback Environment in Acoustic Crosstalk Cancelation

Abstract

Audiovisual immersive interfaces are growing in popularity in today’s world. For the audio part, the "immersive experience" can be accomplished by means of systems that are able to surround a listener with sounds coming from arbitrary locations. When that is done with a set of loudspeakers, the performance of the acoustic crosstalk cancelation process is essential for achieving a convincing immersion. This thesis aims to improve a previously proposed beamforming-based crosstalk cancelation system that uses a linear array of loudspeakers and to verify its performance through simulation. As the original beamformer employed a point-source model for the loudspeakers, here we investigate the effect of loudspeaker radiation properties on the performance of the system and how this contribution departs significantly from that of the point sourcemodel. It is demonstrated that the measured channel separation between the listener’s ears increases when the actual loudspeaker directivities are taken into consideration in the beamformer design. The improvement is mainly noticeable in the frequency range of 1-2 kHz and is globally approximated to 3 dB over the frequency range where beamforming is applied. This thesis also investigates the perceptual effect of different reflecting surfaces that are apparent in the reproduction environment on binaural audio content that is presented through that system. A user study shows that as reverberation from the playback environment increases, the general perception is more pleasant, the impression of space is expanded and feels more real, the front-back confusion is mitigated and even a strong lateral reflection does not weaken the localization cues in a significant way