Sound field control theory and design for the creation of personal sound zones

Sound zone reproduction methods enable the division of an acoustic space into multiple zones. This scenario enables delivering multiple audio contents simultaneously in spatially separated zones. The ability to create individual sound zones without requiring them to be physically isolated is beneficial in many areas such as shared offices, transportation, movie theatres and home entertainment systems. A challenge in sound zone reproduction is to maintain satisfactory performance in reverberant environments. Sound zones are created by filtering loudspeaker signals to produce constructive and destructive interferences in different spatial regions. However, strong reflections from walls may degrade the personal audio listening experience. Hence, creating sound zones often requires measuring the sound field of a region of interest to devise an accurate corrective measure.

This thesis first defines the scope of our investigation and provides the mathematical foundations of acoustics theory. Then, various control strategies are reviewed. They allow the creation of sound zones based on physical models or on measurements of the sound field. To assess the performance of these methods, a finite element analysis with various boundary conditions is employed. The results indicated degradation in the sound zone performance when physical-model-based control strategies were used to create sound zones in a reverberant environment. By contrast, sensing-based control strategies showed improved performance in a reverberant environment. The drawback of sensing-based control methods is their dependence on the quantity and accuracy of measurements of the sound field. To overcome these limitations, different approaches to characterise the sound field were reviewed. The aim was to estimate the sound field at new positions without relying on additional physical measurements. In a finite-element simulation framework, sparse representations of the sound field were employed to calculate estimates through optimising an inverse problem. The accuracy of the estimates was then evaluated and correlated with the conditioning of the inverse problem. The results showed that sparse representations of the sound field allowed accurate estimation. This approach could improve the performance of the sound zone reproduction in reverberant conditions, using a significantly reduced amount of physical measurements. As a proof of concept, the presented approaches were experimentally validated with a real-world sound zone setup developed for the study. It was again found that sparse representations provided an accurate estimation of the sound field of the acoustic system, especially up to 500 Hz, using only five measurement positions. The analysis also indicated that the estimation algorithm could be more beneficial compared to cases for which no estimation technique was used. The performance of the sound zones was improved when the normalised mean square error of the estimation was below -15 dB. Overall, the results indicate the validity and potential benefits of using sparse-model-based sound zone methods.

Future research could formulate an inverse problem to derive the optimal measurement positions to consider when designing sound field sparse models. The sparse method framework presented here could also serve as basis for developing an adaptive room impulse response compensation system.