On the optimization of constrained channel codes

Abstract

A variety of techniques have been proposed for the construction of codes for input restricted channels, including variable length codewords, codes based on partitioning the state successor trees, and methods based on state splitting. In this paper, new methods that avoid exhaustive search are proposed for partitioning state successor trees into subtrees termed independent paths that are used for the coding. First, the approximating eigenvector algorithm is used to determine the weights of the states. These weights are then partitioned into integer parts, which are used to form the independent paths (IP's). Consistency of the weights alone is checked in the first phase of the algorithm, and in the second phase, the partitions are used to form the IP's and to allot information symbols. These methods can also be used to determine the sequence of splits needed for the state splitting technique.