ADAPTIVE BINARY DETECTORS

Abstract : A sequential binary detection problem is considered in which it is required to detect the presence of absence of a stochastic signal in each member of a sequence of observations perturbed by additive noise. The problem is formulated so that the decision rule for the (k1) observation depends on a memory function of the previous k observations. The resulting decision rule which minimizes the probability of making a detection error is complicated by the fact that it is not known which of the past observations actually contain the unknown signal. A timevarying linear predictor of the unknown signal is introduced as the memory function, and a binary detector with a variable structure depending on the linear predictor is discussed. (Author)