Viterbi decoding method

Abstract

A Viterbi decoding method, in which consecutive input data bits Q.sub.6n .about.Q.sub.1n are first converted into consecutive code bits Q.sub.6n .about.Q.sub.3n Y.sub.2n Y.sub.1n Y.sub.0n at a time instant n by use of a Trellis encoder. Then, a branch metric, which is a minimum Euclidean distance between a reception signal point and each of signal points (namely, Trellis codes Q.sub.6 .about.Q.sub.3 Y.sub.2 Y.sub.1 Y.sub.0) in a signal space diagram, corresponding to each of eight states 0 to 7 and a corresponding signal point are obtained. Further, a sum of the branch metric corresponding to each state and a corresponding path metric at a time instant (n-1), which is a sum of corresponding branch metrics since a time instant 1 to the time instant (n-1) is calculated. Then, the path metric corresponding to each state is updated by replacing it with the calculated sum of the corresponding branch metric and the corresponding path metric obtained at the time instant (n-1). Subsequently, a minimum one of the path metrics is determined as the latest part of a surviving path. Moreover, the surviving path is traced back to a signal point Q.sub.61 .about.Q.sub.31 Y.sub.21 Y.sub.11 Y.sub.01 at the time instant 1. Thus input data bits Q.sub.61 .about.Q.sub.31 Q.sub.21 Q.sub.11 at the time instant 1 is restored. In case of this method, the signal points are classified according to three bits Y.sub.21 Y.sub.11 Y.sub.01 into eight groups. Further, one of the groups is determined as a reference group. Furthermore, tables representing the signal points of the other groups are calculated from a table representing the signal points of the reference group.