Custom Architecture for Immersive-Audio Applications

Abstract

In this dissertation, we propose a new approach for rapid development of multi-core immersive-audio systems. We study two popular immersive-audio techniques, namely the Beamforming and the Wave Field Synthesis (WFS). Beamforming utilizes microphone arrays to extract acoustic sources recorded in a noisy environment. WFS employs large loudspeaker arrays to render moving audio sources, thus providing outstanding audio perception and localization. Research on literature reveals that the majority of such experimental and commercial audio systems are based on standard PCs, due to their high-level programming support and potential of rapid system development. However, these approaches introduce performance bottlenecks, excessive power consumption and increased overall cost. Systems based on DSPs consume very low power, but performance is still limited. Custom-hardware solutions alleviate the aforementioned drawbacks, but designers primarily focus on performance optimization without providing a high-level interface for system control and test. To address the aforementioned problems, we propose a custom platform-independent architecture that supports immersive-audio technologies for high-quality sound acquisition and rendering. An important feature of the architecture is that it is based on a multi-core processing paradigm. This allows the design of scalable and reconfigurable micro-architectures, with respect to the available hardware resources, and customizable implementations targeting multi-core platforms. To evaluate our proposal we conducted two case studies: We implemented our architecture as a heterogeneous multi-core reconfigurable processor mapped onto FPGAs. Furthermore, we applied our architecture to a wide range of contemporary GPUs. Our approach combines the software flexibility of GPPs with the computational power of multi-core platforms. Results suggest that employing GPUs and FPGAs for building immersive-audio systems, leads to solutions that can achieve up to an order of magnitude improved performance and reduced power consumption, while also decrease the overall system cost, when compared to GPP-based approaches.