A new approach to sequential detection

Abstract

The known suboptimal sequential detection methods need a minimum-phase signal as input for achieving good performance. As a consequence the received signal must be first transformed to minimum-phase by an adaptive prefilter if simplified detection methods are applied. This roundabout way is avoided in a novel algorithm for sequential detection, where the metric calculation itself is fitted to the special situation of the compare-and-select procedure. The detection method uses two different metrics, one for comparing paths, which have merged in a common trellis-state, the other for comparing paths which have just emerged from a common state. In this way the usual adaptive prefilter can be saved, because the proposed detector does not require a minimum-phase signal at its input. Simulation results show excellent management of distorted and noisy signals.