Accession Number : AD0427881


Title :   CANONICAL FORMS FOR ERROR PROBABILITIES IN BINARY BAYES DETECTION,


Corporate Author : AIR FORCE CAMBRIDGE RESEARCH LABS L G HANSCOM FIELD MASS


Personal Author(s) : Middleton,David


Report Date : Nov 1963


Pagination or Media Count : 32


Abstract : Canonical relations are obtained for the error probabilities and distribution densities of optimum binary detectors in the 'composite' situation involving detection in noise of signals of one general class (S1) vs signals of another general class (S2). These canonical relations are expressed by appropriate distribution densities of the associated 'on-off' cases S1+N vs N, and S2+N vs N, by which it is then often possible to carry out explicit and exact evaluations for the error probabilities in the corresponding 'composite' cases where a direct approach fails or is extremely difficult. The results are general and apply for continuous as well as discrete sampling, for nonnormal, nonadditive, and nonstationary noise processes, as well as for more familiar normal, additive, and stationary cases. Specific use of these canonical relations is illustrated by several examples. (Author)


Descriptors :   *INFORMATION THEORY , DETECTION , NOISE , SIGNALS , PROBABILITY , NARROWBAND , BACKGROUND


Distribution Statement : APPROVED FOR PUBLIC RELEASE