The effect of non-equiprobable source symbols on maximum-likelihood Viterbi decoder performance

The effect of non-equiprobable source symbols on maximum-likelihood (ML) Viterbi decoder performance is investigated. Performance impact is determined by comparing the bit error rate (BER) of the Viterbi algorithm using an ML metric to one based on a joint, maximum a posteriori estimator of the source symbol probabilities and the optimum path through the code trellis. The algorithm is simulated using hard decision decoding and upper bounds on its performance are presented. Simulation results reveal that significant BER degradation may be experienced at low to moderate E_b/N₀, ratios if the source symbol probabilities are skewed from the uniform case