Joint robust decoding and parameter estimation for convolutionally coded systems impaired by unknown impulse noise

By assuming the impulse statistics, channel coding has long served as an effective tool against fairly frequent occurrence of strong impulses, which impedes the guaranteed quality of service of communication systems. Nevertheless, impulse statistics are generally hard to be modeled and likely to be time-varying, posing a great challenge to system design. Our previous work reveals that, without assuming any impulse noise model, the metric erasure Viterbi algorithm (MEVA) can achieve the performance derived from the (benchmark) maximum likelihood decoding in single-carrier convolutionally coded system subject to memoryless strong impulses, provided that the clipping threshold is judiciously selected and the power strength of the background noise, σ² is known. In an attempt to alleviate the assumption of using σ², this paper proposes the iterative MEVA (IMEVA): the information regarding the indicator sequence (which is induced by the MEVA) is leveraged and passed to the σ²-estimator for refining the clipping threshold estimate at the next iteration. Regardless of the impulse noise model encountered, simulation results show that the IMEVA performs remarkably close to that of the Baum-Welch algorithm, which, at the cost of receiver complexity, relies on knowing the underlying impulse noise model.