Threshold detection in generalized non-additive signals and noise

The classical theory of optimum (binary-on-off) threshold detection for additive signals and generalized (i.e. nongaussian) noise is extended to the canonical nonadditive threshold situation. In the important (and usual) applications where the noise is sampled independently, a canonical threshold optimum theory is outlined here, which is found formally to parallel the earlier additive theory, including the critical properties of locally optimum Bayes detection algorithms, which are asymptotically normal and optimum as well. The important Class A clutter model provides an explicit example of optimal threshold envelope detection, for the non-additive cases of signal and noise. Various extensions are noted in the concluding section, as are selected references