Multiple description data compression is a new signal processing paradigm which has significant applications in packet transmission of video, image and speech signals. The idea is to construct several descriptions of the source sequence (unlike standard compression in which only one description is constructed). Each description should carry sufficient information to reconstruct the source signal with acceptable quality. In addition, the combined information in the descriptions should be sufficient to reconstruct the source sequence with high quality.
The first part of this talk will provide an overview of research in multiple description data compression. The second part will focus on recent results. We consider the problem of designing a multiple description quantizer which uses a root lattice as a codebook. In order to design such a quantizer we must first solve a lattice labeling problem. We will focus on this labeling problem in the second part of the talk.