Multiple Description Coding for Packet Video

Signal Compression Laboratory Research Project

 

Researcher: Prashant Koulgi
Faculty: Prof. Kenneth Rose
Research Focus: Recently, diversity techniques to increase the robustness of communication systems and storage devices have been receiving attention. Examples of such systems include packet networks, wireless systems using multi-path and Doppler diversity and Redundant Arrays of Inexpensive Disks (RAIDs). Multiple Description Coding (MDC) has been shown to be an effective technique for robust communication over such erasure channels.

One of the main topics of research in MDC has been the direct design of Multiple Description Quantizers. Current design techniques exist only for scalar quantizer design under some constraints. Even these design techniques suffer from the problem of poor local minima. Our work is focused on the design of unstructured Multiple Description Vector Quantizers (MDVQs) for diversity systems employing two effective channels. Currently we are investigating the use of the Deterministic Annealing (DA) approach for this problem. Our approach avoids many poor local minima, and out-performs existing techniques for the special case of scalar quantizer design. Further, this approach can also be used for the design of vector quantizers without any modifications.

Presentation Multiple Description Quantization by Deterministic Annealing