A Burst Error Model Using Interleaved Bernoulli Processes with a Markov Property.

摘要

A model of burst errors from a real-time error bit stream of up to 20 million bits per second is sought. Noise models and channel models are considered. The choice is a two-state Markov channel model in which the states each generate independent Bernoulli random processes. One Bernoulli process produces ones (errors) at a high rate, simulating burst errors. The other process produces ones at a low rate, simulating random errors due to background noise. The transition probabilities determine the average length of the bursts and of the gaps. Relative frequency estimates of the probabilities of certain sequences of one-bits from real data are related to estimates of the model parameters, so relative frequencies provide a basis for fitting this model to real channels using observed error sequences. an equation for the number of errors in a block of bits is developed in terms of the model parameters. Burst probabilities can be predicted based on this equation. The model was tested using computer simulation. Some discussion is devoted to how this burst-error model can be implemented in an actual device to provide real-time channel characterizations. This model aids in the selection of an error correction code. (Author)