Non-linear codes and concatenated codes for unequal error protection

We will present two different error control schemes for unequal error protection. Based on a general lower bound on the word error rate (WER) for the coded transmission of redundant data over discrete memoryless channels, the distance profile of the optimum error correcting code can be derived. This distance profile may serve as a basis for code construction as it is done by an approach applying a systematic computer search to the code construction which yields a non-linear code. The construction as well as decoding of the code is quite complex, hence, the length of the codewords is limited. The other approach combines concatenated Reed-Muller codes with iterative `turbo´ decoding techniques resulting in a powerful coding scheme suitable for unequal error protection (UEP). This facilitates the construction of very long codes from simpler shorter codes, since the decoding complexity depends only on the applied component codes. But compared to the first method the codes are less flexible as far as the design of the distance profile is concerned