Low-complexity Viterbi decoder for convolutional codes in Class-A noise

The design of a simplified Viterbi decoder for signals in Middleton Class-A noise is considered. The conventional Viterbi decoder, with a branch metric optimized for Gaussian noise, performs poorly in the Class A noise. The optimal maximum likelihood (ML) branch metric is difficult to simplify due to the complexity of the probability density function of the noise. There are different alternatives to design low complexity Viterbi decoders which are based on simplified models of the Class-A noise. Furthermore, a nonlinear preprocessor has been proposed to improve the performance of the Gaussian Viterbi decoder in Class-A noise by using a simplified expression of the probability density function of the noise. In this paper, we propose different approach to design the Viterbi decoder with simple linear branch metrics by using a simplified linear approximation of the log likelihood ratio. The proposed approach results in near-optimal performance with low complexity