A musically motivated approach to spatial audio for large venues.

摘要

Spatial audio refers to those aspects of sound reproduction associated with the perceived direction and expanse of sound images. The vast majority of research on spatial audio considers a single listener, ideally positioned in the listening area among the speakers. From that position, spatial effects can be quite convincing. However, this approach cannot be extended to large or public spaces, or to non-centered listeners. Many existing solutions are limited to cinematic special effects or have extensive hardware requirements.;This dissertation presents a new approach to large-venue spatialization, particularly appropriate for musical and creative artistic works. Digital signal processing (DSP) is used to explicitly control the region of focus in the listening area, and hence which audience members are receiving spatial cues. By moving the target area over time, the entire audience can receive maximally effective spatial audio at some point. Target location processing is guided by consideration of the entire soundfield. The emphasis is not on absolute localization accuracy but on potential aesthetic uses. The system is not limited to localization cues; decorrelation effects are treated with equal attention.;A spherical head model is used, both for synthetic directional cues and for predicting the physical acoustic response. Targeting of listener locations is performed using several methods, including crosstalk cancellation. This work also makes the novel assertion that overhead loudspeakers create an optimal soundfield, both by maximizing the target area and by minimizing undesirable properties elsewhere. Evidence is provided to quantify and support this claim.;A historical review of prior approaches to spatial audio in large venues is given first. The relevant literature on spatial hearing perception is then summarized, with emphasis on large-space loudspeaker reproduction. Next, the signal processing techniques used for implementation are discussed. Evaluation is done primarily through simulations of soundfields using recorded Head-Related Transfer Functions (HRTFs). In addition, listening demonstrations were conducted to confirm the correct operation of the system.