Spatial audio rendering with flat-panel loudspeakers

Thesis (Ph. D.)--University of Rochester. Department of Electrical and Computer Engineering, 2018.Flat-panel loudspeakers have existed for nearly a century, with recent advances combining signal processing and structural acoustics to improve audio quality. The incorporation of signal processing allows these advances to be adapted for spatial audio purposes. Methods for rendering spatial audio with flat-panel loudspeakers are investigated in this thesis, beginning with a brief history of spatial audio rendering and the use of multi-actuator panels for wave field synthesis. Mechanical models are developed to describe the bending motion of thin panels excited by both moving-coil, and piezoelectric actuators. A modal diversity metric is introduced to characterize naturally occurring vibration localization on panels excited by a single actuator. The radiation characteristics of these localized regions are found to approximate point sources if the panel is heavily damped and excited well below its coincidence frequency. The use of actuator arrays to create localized panel vibrations is explored in the regime where panel vibrations are not localized naturally. The magnitude and phase parameters of each actuator are tuned to selectively excite specific linear combinations of panel bending modes, the superposition of which forms a band-limited spatial Fourier series approximation of a localized vibration region. Methods are developed to optimize the placement of actuators to most efficiently address a given set of modes, equalizing the acoustic response of each localized source, and dynamically moving the localized source to different regions of the screen in real-time. Techniques for predicting loudspeaker preference ratings are applied to several different panel loudspeakers to determine perceptually how the array-driven approach differs from traditional design methods. Array-driven panels employing modal crossover networks and equalizations consistently score higher ratings than panels excited by a single force actuator with low modal diversity. Finally, a system is introduced for object-based audio rendering on near-field displays for use with encoding schemes such as MPEG-H 3D Audio