Technical Report: Project PERCEPT MEDIA, July 1999

Project Title: Scalable Stream Coding for Adaptive Foveation Enhanced Percept Multimedia Information Communication for Interactive Medical Applications
Organization: Kent State University
AO Number: H491
Contract Number: F30602-99-1-0515
Start Date: 15 JUN 1999
End Date: 15 JUN 2001
Principal Investigator
Javed I.     Khan
Address: 233 MSB
  Kent State University
City,  State   Zip: Kent,   OH     44242
Phone: 330-672-4004
Fax: 330-672-7824
Level Of Participation - Billed: 50% 
Level Of Participation - Unbilled: 50% 
Financial POC
City,  State   Zip:
Project URL:
Objective: Objective: With its popularization, Internet is rapidly entering into an era of sights and sounds and is moving beyond the traditional packet data communication. Unfortunately, the popularization and consequent expectations from the global computer network have thus far outpaced the technological development. The current infrastructure is capable of connecting an estimated 2 billion people worldwide. It is expected to double every 25 years. Assuming just 15 minutes a day of multimedia integrated communication, per person with acquaintances worldwide (with inverse geometric distribution between duration with hops), it will create a sustained traffic of 3-10 Gbps per local area network, with 100 times data compression technology. Multimedia communication offers some hard challenges to the existing network infrastructure. Multimedia data forms differ from the legacy data, which the current computer communication networks are designed to carry. The two key differences are (i) they exhibit stricter temporal dependency. For seamless communication it requires a significant amount of data to be communicated with low delay and jitter. (ii) They can operate on less than perfect data integrity. Some part of the communicated data requires accuracy, while for bulk of it the accuracy requirement is less stringent. More precisely, the 'quality' of transmission is not uniform for all it's component. Legacy computer communication networks, protocols, instruments, algorithms have been designed for carrying asynchronous data with almost utmost perfection. Thus, multimedia data creates some awkward difficulty for legacy networks. The transmission schemes that have been developed over the years have also been trapped into some limitations. From the application point of view, the first generation transmission schemes have been designed without anticipating much sophistication from the networks; consequently they lack flexibility and adaptability. For example: Data compression algorithms mostly depend on source side analysis to remove information redundancy. These techniques roughly offer 10-100 times compression [5]. Recent trends show that current compression techniques, which are based on source analysis, are approaching, a form of theoretical limit. Also, the multimedia coding schemes have been designed to target receivers of almost similar capabilities. These schemes offer limited flexibility to adjust to the diversity of the receivers and their working environment. As the computer communication network has grown to global proportions, diversity and dynamic variations are not exceptions but a rule. Adaptation ability is not only vital for the future complex networked applications, but it will also be a key instrument for improving overall resource utilization in a large, complex, and shared environment. Current transmission schemes are ill equipped to adapt. The emerging ANET concept has the potential to fundamentally benefit multimedia communication in overcoming both of above the limitations. It can offer a dynamic 'sink-side' feedback based data-compression. It can also offer a new range of dynamic receiver state -based adaptability and scalability to data-streams. The key empowering factor here is the new capability of midstream 'interaction' between the downstream and upstream information flows of a single communicating system. The objective of the proposed research is to demonstrate how the intelligence and flexibility of the powerful concept of active networking can help multimedia by offering a self-configurable and adaptive scalable data stream with dynamic interaction from the receiver in the form of perceptual feedback. The objective is to (i) develop an adaptive transmission scheme which will utilize the dynamical perceptual fovea information at the receiver to maximize user's perceptual satisfaction, (ii) benchmark it over proposed ANET platforms and evaluate the performance, and (iii) identify the nature of further development of the ANET platforms as well as the scalable coding technology.
Approach: Approach: The objective of this research is to show how percept multimedia applications can dynamically adjust to the varying network, link, protocol and environmental variations that may exist at the communication system including the application-end-points, as well as how sink-side feedback can improve coding efficiency. The proposed approach is based on (i) the newly emerging active networking platform and (ii) a novel priority based information-coding theory for percept data. The scheme will use the existing state-of-the-art compression algorithms that can take full advantage of source side statistical and perceptual redundancy extractions. However, a novel packetization and composition technique based on prioritization theory will be used which will allow receiver triggered controlled reduction of information. The actual reduction will be performed at the network elements upstream or at the source. This scheme will be implemented by deploying application-level stream and session specific adaptation protocols via active networking. Based on the priority scheme a novel packet structure will be defined. These structured containers will have jettsionable compartments (JET packets). Based on the priority-model, application streams will be packed into the JETs. An interaction model will be developed by which JETs can be trimmed dynamically midstream at the active nodes. The trimming information will be dynamically fed to the ANET distribution nodes from downstream. The trimming streams will originate at sink-side containing the viewer's dynamic perceptual-feedback, dynamic receiver condition and downstream network condition. All three of these up-flowing information streams will interact with the down-flowing data stream at designated ANET distribution nodes. The ANET distribution nodes will maintain application state information. The general schemes will be applicable not only for video, but for other data-forms (such as audio, video, 3D animation, synthetic model that generates it, etc.) which are in the path or eventually terminate at a perceptual sink (such as human eye). For concept demonstration, the developed technique will be applied to 'activate' the multimedia percept data-paths of a state-of-the-art distributed radiation treatment planning system. The system will be tested on several of the experimental ANET platforms.
Recent Accomplishments: new start.
Current Plan:
  • Development of a video code segmentation and priority model. 
  • Design of the scalable coding and packing schemes. 
  • Development of the active node state engine and interaction model. 
  • JET Packet dynamic trimming demonstration. 
Technology Transition: new start.
Comments / Questions / Anything else you need: