Skip to main content

Advanced Multimedia Parallel Accelerator


Recent multimedia workstations and networks are capable of handling complex information formats like sound and video, making it practical to implement applications like CSCW and video-conferencing. However, the requirements of high quality images for broadcasting and some of the advanced applications like Virtual Studios, tele-presence and Virtual Reality require much more powerful facilities than are available today. In order for some of the important advances to be achieved in the above fields to reach the mass consumer markets, the underlying costs of such facilities need to be reduced considerably.
A number of ACTS projects, in particular in the Multimedia Services Domain, include hardware support targeted to specific requirements. It is important that a consistent approach within a framework exists so as not to dissipate effort in a number of specialised hardware implementations. In fact MPEG-4 is attempting to define a comprehensive overall framework with a whole range of profiles and levels as already defined, so as to standardise a number of state of art image coding and multiplexing techniques. As such standards are accepted for new television broadcasting services and as the basis for multimedia services via telecom networks, it becomes important that architectural support is developed to implement these.
The detailed specification for an Advanced Multimedia Parallel Accelerator (AMPA) proposed to be built by the Consortium has been completed and presented at the ACTS Concertation meeting. This platform provides a high performance architecture designed for processing real time video and data streams and combinations of these for advanced multimedia applications. AMPA can be implemented at multiple levels, allowing a range of costs and performances to be based on common software and scaleable hardware architecture. The specification includes a description of the range of AMPA architecture configurations, the component technologies, an evaluation of interconnection networks, and systems software aspects.
Work is in progress in completing a survey of the novel architectures for object-oriented multimedia with special emphasis to MPEG-4 within ACTS, to place AMPA within wider context. For this purpose working relationships have been established with ongoing architecture work within the existing projects related to MPEG-4 such as EMPHASIS, MOMUSYS, and SCALAR and contacts have been established with MIRAGE, VANGUARD, AURORA, and others.
Work has started in investigation concurrence in MPEG-4 encoding and decoding, in particular in VOP processing, splitting VOPs for parallel processing and load balancing.
Expected Impact
The unified architectural framework based on mainstream workstations enhanced by the state of the art yet low cost hardware/software is to provide a common platform for bringing to the market some of the work undertaken by ACTS projects in the Interactive Digital Multimedia Services domain, in particular related to MPEG-4, and thus making the output of such projects much more readily available, providing an opportunity of market lead to related European companies.

Main contributions to the programme objectives:
Main deliverables
Feasibility study for the applicability of an advanced multimedia parallel processor to a number of specific applications developed within ACTS projects
Contribution to the programme
Contributed to lowering the entry cost of advanced multimedia systems such as those based on MPEG-4

Technical Approach
To meet the stringent real-time video processing requirements of a number of ACTS projects, in particular in Domain 1. Interactive Digital Multimedia Services, this feasibility study will investigate the application of an Advanced Multimedia Parallel Accelerator (AMPA) to a number of specific applications within those projects. The objectives are for low cost yet state-of-the-art AMPA configurations which can be plugged into mass produced PC multimedia workstations to provide a viable platform for a number of key applications to reach wider user markets. The technical approach is based on the use of an intelligent frame buffer with standard interfaces which can be directly connected to video devices (cameras, monitors, ...) combined with scaleable parallel processing capability for numerically intensive algorithms to satisfy the underlying requirement for real time working. The concurrent use of the multiple clusters provides a scaleable parallel architecture particularly suited for real-time video processing.
Summary of Trial
The work of this project is such that no practical experimentation is foreseen.
Key Issues
The requirements of high quality and low costs, we believe, can be met by add-on specialised hardware to the mass produced multimedia workstations to provide the standard interfaces required by TV cameras, monitors, etc., to handle TV quality video sequences in real-time, and to include enough scaleable processing power through parallelism for numerically intensive algorithms. The underlying requirement for on-line 'live' working is the ability to handle all the operations in real-time and for post-processing in minutes or hours rather than days and weeks

Funding Scheme

CSC - Cost-sharing contracts


University of London
Mile End Road
E1 4NS London
United Kingdom

Participants (2)

DVS Digitale Videosysteme GmbH
Loughborough Sound Images (LSI)
United Kingdom