Multiscale unit-memory convolutional codes

In this study, authors propose structure to unit memory (UM) convolutional codes that greatly simplify their decoding when using sequential or suboptimal L-decoders. The multiscale structure proposed in this study preserves the word-oriented nature of unit memory codes (UMCs), but allows for processing the code in blocks of sizes less than the memory of the code with non-Viterbi-type decoders. At one end of the multiscale structure are generic UMCs. At the other end are structured systematic UMCs that can be decoded with the same ease as conventional multimemory convolutional codes with sequential or L-decoders. As a special case, the authors show quick-look parity check systematic codes of arbitrary rates that are better than previously known optimum minimum distance systematic convolutional codes. These codes are derived from block codes and the convolutional code can be decoded using the syndrome decoder of the underlying block code. The authors present simulation results and analysis to support the proposed multiscale UM convolutional codes.