USE OF A NOISE CLASSIFIER IN DETECTION OF A SIGNAL IN NOISE.

A noise classifier can be used in systems for detection of a signal in a background of some one of a set of noises when noise alone is present in the initial observation interval. Cases are examined in which the a priori information is given completely and in which there is a priori indeterminacy. It is shown that use of a classifier reduces the problem of testing compound hypotheses to problems of classification and testing of simple hypotheses. A suboptimal decision rule for the detection system with a classifier proves to be simpler for synthesis and realization than the optimal rule for testing compound hypotheses. Estimates are obtained for the ″losses″ in solution quality with use of a suboptimal rule.