Three types of robust maximum-likelihood Viterbi decoder

Abstract

We propose three new types of maximum-likelihood Viterbi (1967) decoders (for convolutional or trellis codes) with each under the hypothesis of logistics, Laplace, of Cauchy noise distribution respectively. Unlike a Gaussian ML Viterbi decoder, the three newly-proposed Viterbi decoders exhibit a valuable robustness property, i.e. insensitivity to deviations from the assumed noise distribution.