M-path decoder with algebraic metric calculation

Abstract

The authors propose a modification of the M-path convolutional code decoding algorithm in order to improve its performance. Due to nonexhaustive exploration of the trellis, the M-path algorithm is a suboptimal decoding algorithm. On the other hand, this elimination of certain paths from the trellis reduces its complexity in comparison with the Viterbi algorithm. The proposed modification adds to the traditional calculation of the branch metric an algebraic metric representing an estimate of the number of future symbol errors. By this new metric the performance of the M-path decoder is improved by about 0.5 dB.