The VITERBI-Algorithm for impulsive noise with unknown parameters

We propose a modification of the well-known Viterbi algorithm (VA) for communication channels distorted by impulsive instead of the often used Gaussian noise. Here we assume that the parameters - eg, the moments - of the noise are unknown. Instead of applying a recursive solution by repeated execution of the VA we directly embed the estimation of the unknown parameters into the structure of the VA itself. Such an approach is called per-survivor processing (PSP) which provides a general framework for the approximation of maximum likelihood sequence estimation (MLSE) whenever the presence of unknown quantities prevents the precise use of the classical VA. In addition, the classical VA is modified so that it works optimally for some kinds of impulsive noise. We show by means of the modified VA, that the bit-error rate can be substantially decreased. In other words, only with minor technical modifications by minimizing an adequate nonlinear norm, the transmission becomes more reliable compared to the usual Euclidian norm minimized by the conventional VA