Design of a real-time GPU accelerated acoustic simulation engine for interactive applications

room acoustics;audio;Digital Signal Processor (DSP);Graphics Processing Unit (GPU);parallel;video games;simulation

Cheng, Zuofu

关键词: room acoustics;    audio;    Digital Signal Processor (DSP);    Graphics Processing Unit (GPU);    parallel;    video games;    simulation;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/50364/Zuofu_Cheng.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Current-generation video game applications use sampled recordings along with positional audio and modular effects to create the audio experience. While computationally efficient, this presents multiple immersion and game play limitations. We propose instead a system for simulating room acoustics in real-time based ongeometry and material descriptors. This system uses commodity graphics processors (GPUs) and leverages the parallelism inherent in a ray-based acoustics model to compute the reflected and transmitted acoustic response from a number of sound sources to a stereo listener in arbitrary triangle-based geometry. Our goal is to describe a complete acoustic simulation engine which requires little additional information beyond what is already available to the game engine andgenerates realistic and immersive game play audio. First, we present background on current generation video game audio as well as advancements in architectural room acoustic simulation. This survey demonstrates both the limitations in current generation applications, as well as establishes computation and software design constraints which a replacement systemmust satisfy. Chapter 1 will also discuss the theoretical justification for ray acoustic methods, as well as limitations in certain scenarios. We will present this in the context of current generation architectural acoustic simulation software, which largely represents the state-of-the-art understanding of the underlying acoustics which govern sound propagation in spaces. Finally, Chapter 1 will also introduce useful concepts and terminology which will justify some of the engineering decisions made in the remainder of the text.The remainder of the dissertation will discuss the technical construction of the system. Specifically, it attempts to justify the engineering decisions and trade-offs which were made in order to allow for real-time performance. We will discuss the organization of the GPU family for which the system was designed, and how the problem was tailored to fit within the constraints of the computational architecture. Optimizations specific to the architecture used will be discussed here, as well as their subjective effects on sound and the experience quality.