Design methods for VQ by linear mappings of block codes

Vector quantization (VQ) is one of the most important building blocks in modern source coders. Both speech and image compression systems are strongly dependant on VQ. When increasingly exploiting the compression capabilities of VQ, we normally have to pay a price in terms of problems with channel robustness and complexity. We have proposed linear mappings of block codes (LMBC) as a technique well suited to deal with some of these problems. As the scheme includes both a block code and a mapping, it is not obvious how to do the design. We demonstrate several design methods that work well and discuss their properties