A statistical theory of automatic observing systems

An observation on a known and partially unknown signal environment is formulated mathematically. The detailed structure of an observing system capable of recognizing "new" and "old" signals is obtained along with mathematical expressions for predicting system performance. The theory is applied to several well-known problems in signal recognition. The results of an experimental study concerned with the asymptotic performance measures are presented. The logical design of a feasibility model based on the theory is discussed.