Complexity Optimization Strategies for Adaptive Multimedia Receivers

University Of California-Los Angeles, Los Angeles CA

Investigators

Abstract

This project is premised on the belief that a fundamentally different approach is necessary for designing and implementing multimedia compression, protection and transmission algorithms and systems for resource-constrained networked devices. The main idea is that instead of considering multimedia algorithms as a given and adapting the implementations reactively to match the requirements of these algorithms, proactive joint optimization of the algorithms, their parameters, and the implementation task partitioning is performed. The system is therefore able to maintain operation at a point in rate-distortion-complexity space in accordance with a constellation of factors and constraints including distortion tolerance, power, platform architecture characteristics, network behavior, and critically, data attributes. A global optimization framework and associated algorithms that consider the data-specific impact on the system factors listed above, allowing cost functions describing these tradeoffs to be formally described and then optimized as appropriate to different devices, networks and applications will be developed. Specifically: 1) complexity description methods and models, 2) data-aware channel decoding methods that combine feedback from the decoder and system knowledge to adapt the processing, 3) joint source-channel coding methods that exploit error concealment in combination with knowledge of the time- and content-varying sensitivity of multimedia data to errors, and 4) optimized resource management. The broader impact of the project is in the area of educating next generation students and industry partners in optimized resource-aware system design. Developing formal methods, algorithms and models will lead to improved multimedia performance over existing resource-constrained systems and also provide valuable insights into the design of next generation multimedia compression algorithms that should be complexity scalable, as well as system designs that should be multimedia aware.