Convolutional Error Detection on an Additive White Gaussian Noise Channel

Abstract

Concatenated coding schemes involving a convolutional inner code and a block outer code have occasionally been used in communication systems that are very intolerant of errors. In these schemes the vast majority of channel errors are corrected by the convolutional decoder while the block outer code is used to detect convolutional decoder errors. Block code words containing detected errors are erased. Soft decision Viterbi convolutional decoders operate by comparing path metrics and selecting the path with the largest metric (the maximum likelihood path). There is a substantial amount of information in the path metrics that is not used in this pick-thelargest decision. It is proposed that some of this information be used in a probabilistic decoding error detection scheme. Such a detection scheme would obviate the use of the block outer code. The result is a bandwidth savings at the cost of some additional processing of the convolutional code metrics.