Abstract
One of the characteristics of the Viterbi algorithm is that the errors at the output tend to group together in clusters called bursts. It is hence useful to know the burst error statistics in designing a communication system which employs the Viterbi algorithm. In this paper we use the method of the hidden Markov model (HMM) to develop a Markov chain model for the burst error statistics of the Viterbi algorithm. Though only the case of Viterbi decoding is considered, the results presented can be employed to other applications of the Viterbi algorithm. One of the advantages of building such a model is that it can be used to generate the output sequence with little cost when compared with the real simulation of the Viterbi algorithm and provide a basis for studying other system parameters. The HMM we develop in this paper generally performs better than the geometric model proposed by Miller et al. and in most cases better than the model by Chao and McEliece, and it requires much less parameters than those of the model by Chao and McEliece for convolutional codes of large constraint length.