Performance of the Viterbi algorithm for interleaved convolutional codes

Abstract

This paper presents a maximum-likelihood decoder for error-burst channels with a very efficient implementation. In particular, the encoder is formed of an interleaved convolutional code with generator polynomials of the type [g/sup (1)/(D/sup I/), g/sup (2)/(D/sup I/)] (for a rate 1/2 coder), where I may assume a very high integer value. The decoder consists of the Viterbi algorithm (VA) optimized for these sparse polynomials. For a given decoded bit-error probability, the required delay time and memory requirements of this approach are more inferior by far than those of the traditional method of interleaving. Moreover, for bursts of an average length less than 1/2, this method provides better performance than the dual-mode burst-error-correcting algorithm.