On error exponents for a variable size list decoder using the Viterbi algorithm with likelihood testing

Abstract

SUMMARYThe likelihood ratio (LR) decision criterion [3] employed by Yamamoto and Itoh's decision feedback (ARQ) scheme [1] is known to be highly compatible with the Viterbi algorithm (VA) due to the fact that it requires a small amount of computational effort. In addition, Forney has derived an extension of the coding theorem to the variable size list decoder for maximum likelihood decoding of block codes. In this paper we propose an algorithm for extending LR-based variable size list decoding for decoding convolutional codes using the Viterbi algorithm; assuming a discrete memoryless channel we also derive a lower bound on the decoder's error exponents.