Characterization of digital channels using hidden Markov models

To characterise the error process in binary symmetric channels with memory, the hidden Markov model allows more powerful modelling than the commonly used Fritchman models. Baum-Welch re-estimation is used to infer the model parameters from error sequences measured over the channel. This method is computationally very demanding for long error sequences, which are necessary when low BER channels are modelled. An efficient re-estimation has computational load directly proportional to the number of errors in the sequence rather than to its length. Channel simulation may be speeded-up similarly