A network flow approach to convolutional codes

Presents an alternative way of looking at convolutional codes as a network of flows. From this association, previous known classes of nonsystematic convolutional codes are reproduced. The authors also present a decomposition of the problem of finding good convolutional codes into well known and structured problems such as the maximum flow problem, the knapsack problem, and the modular representation of block codes. A spanning tree search algorithm is also proposed for finding good codes. The authors consider the class of nonsystematic convolutional codes for which the encoders satisfy an equivalent property related to the feasibility of flow. Finally, extensive tables show the new and the previous known codes.<>