SEQUENTIAL DETECTION II,
JOHNS HOPKINS UNIV BALTIMORE MD CARLYLE BARTON LAB
Pagination or Media Count:
Three problems involving sequential detection are considered in this study 1 canonical forms for weak-signal, or threshold detection 2 conditions for the singularity and non-singularity of optimum detection procedures for normal noise in normal noise and 3 useful experiments involving binary sequential detection, particularly where analytic results are not obtainable. Canonical expressions for locally optimum receivers are obtained for the sequential detection of general classes of signals in general classes of noise when the cost functions are linear in sample-size. It is shown that the necessary and sufficient conditions for singularity and non-singularity for the sequential likelihood discrimination of one normal noise process, N sub 1, vs another, N sub o, is the same as that for the fixed-sample cases, with the additional observation that these results are to be understood as holding on the average, since now the finite interval sample-size, T, is a random variable. A series of experiments involving sequential detection of signals in noise is also described and outlined in general form.