Survivor Error Patterns Based on Structural Properties for Maximum Likelihood Decoding

摘要

Maximum likelihood soft decision decoding can be accomplished by performing a search for an error pattern which, when added to a binary version of the received sequence of symbols, yields the most likely codeword. The measure applied for selecting the best error pattern is the real valued reliability, rather than the Hamming weight employed in case of hard decoding. To facilitate the joint handling of binary and real operations a type of ordering is imposed on the collection of error patterns, whereby certain error patterns called survivors can be specified. The subcollection of survivors contains the best error pattern. This approach yields rather efficient decoding algorithms for the some codes, most notably the Hamming codes and their derivatives. Also, several decoders have been devised and implemented for turbo-like codes whose constituents are Hamming codes. In this paper we consider the extension of the approach to codes with more complicated structure. In particular, the weight distribution of the code will be employed together with reliability considerations in order to specify survivor error patterns.