An efficient reduced state maximum likelihood sequence estimator

摘要

A reduced state maximum likelihood sequence estimator allows the use of improved equalization techniques that provides greatly improved performance for channels with severe attenuation and spectral nulls. The reduced state maximum likelihood sequence estimator retains kn states of a total number of K states, knK, with each retained state having an associated state metric. (J)(kn) new states are determined based on kn previous states and a most recently received sample, using J transitions, J being a less than L, where L is a size of a symbol alphabet. (J)(kn) new state metrics are determined which are respectively associated with each new state. The new state metrics are compared to a threshold and those states whose metric does not exceed the threshold are retained. The reduced complexity of the MLSE allows for the use of partial response equalizers, e.g., a partial response class V (PRV) equalizer. The number of retained states kn varies according to an adaptive threshold which limits the number of retained states between a lower bound and an upper bound. The threshold may be determined according to the metric of the most likely state of the retained states and a selected one of a plurality of transition metrics associated with the most likely state metric.