A Distributed Architecture for Sound Rendering

Audio communication is considered as an important communication channel, after visual communication, to some of the collaborative multimedia applications (CMAs), such as Internet gaming and collaborative architectural design. Unfortunately, real-time synthesis of spatialized sound is difficult due to its high computational costs. Most existing systems are developed for non-real-time applications. Some developed for real-time applications are typically based on multiprocessor architectures. They are for very simple environments and are not scalable. In this project, the researchers propose to develop a distributed architecture for real-time synthesis of spatialized sound. To support collaboration, the new method will support moving sound sources as well as observers. These moving sound sources and observers may represent users of the application that produce and receive sound, respectively. To be able to produce spatialized sound dynamically, the researchers propose to develop three techniques here:a distributed path-tracing method to dynamically allocate path-tracing tasks to multiple servers;a path approximation method to gradually decrease the path resolution of a sound ray as it is being reflected more and more; anda multi-resolution sampling scheme to determine the number of sound paths to be traced from each sound observer.