Presently, an abundance of digital streaming media such as audio and video have become a pervasive and integral part of many applications. The goals of this research project are to study, design and implement a data stream recorder that addresses the challenges inherent in a broad class of innovative applications that acquire, store, manipulate and disseminate digital continuous media (CM) streams in real-time and on a large scale. To this end the research focuses on a unified paradigm that integrates multi-stream recording, retrieval and control in a synergistic manner to provide well-defined services for all media, cognizant of their characteristics such as bandwidth requirements, resolution and modality. The topics of study include novel shared buffer management techniques, statistical admission control algorithms based on a comprehensive random variable model to support both reading and writing of multiple variable bitrate media streams and multi-threshold flow control protocols enabling smooth stream acquisition and playback. Additionally, comprehensive query support is investigated through a declarative language that facilitates the specification of both temporal and semantic relationships in media streams.
The resulting capabilities have tremendous appeal and are expected to have a significant practical impact on a broad range of applications that make use of multimedia sensors, from industrial monitoring - such as intelligent traffic management - to homeland security and entertainment applications. An integral part of the project is the evaluation of the stream recorder design in the domains of information comprehension and immersive environments. The prototype implementation and experimental research of this project will be used as a teaching tool for courses covering multimedia and server technology and will be disseminated via the project Web site at http://dmrl.usc.edu/hydra.html.
