On The Problem Of Optimal Signal Detection In Discrete Time Correlated Non Gaussian Noise PDF Download

A general analysis of the Ternary Class (M = 2): H sub 0: N vs H sub 1: S1+ N vs H sub 2: S sub 2 + N of signal detection problems is is presented, for completely general signals, i.e., both broadband narrow-band, deterministic or random, in generalized (i.e., non-Gaussian) noise, in the limiting threshold regime. This includes optimum threshold algorithms and system performance, as measured by the appropriate error and detection probabilities. The present treatment, however, is subject to the following constraints: (1) independent noise sampling; (2) ambient noise models, i.e., noise independent of the signals; (3) uniform cost functions, e.g., C sub o (> 0) for errors, and C sub 1 = 0 for correct decisions. Under these conditions, only three principal parameters are needed: delta 12, delta 22 = signal detection parameters (= 'output (S/N) 2') and the correlation coefficient P sub 12 (= P) between the two (threshold) test statistics (or detection 'algorithms') Z sub 1, Z sub 2, apart from the a priori probabilities (q, p sub 1, P sub 2) of the presence of noise alone, S dub 1, and S sub 2. Next steps, to extend the treatment to the general case (M = 3): H sub 1: N + S sub 1, vs H sub 2: S sub 2 + N vs H sub 3 : S sub 3 + N, and to include correlated noise samples, are noted ... Ternary detection, Coherent and incoherent reception, Threshold signal detection, Generalized noise.