NONCOHERENT NAR POWER ESTIMATORS FOR ADAPTIVE DETECTION

Quadratic noise-alone-reference (NAR) power estimators are relatively difficult to implement in bandpass systems. This correspondence analyzes NAR power estimators in the form of FIR filters operating on square-law demodulated samples. The standard estimator [formula omitted]s, which in the absence of a signal is unbiased and of minimum variance, has no NAR property. By constraining [formula omitted]s to be nonnegative and unbiased to a symmetrically observed signal, a generalized NAR estimator [formula omitted]GNAR is obtained. It is independent of the signal envelope but requires some pure noise samples to always be present at the input. Therefore, a quasi NAR estimator [formula omitted]qnar is derived to be nonnegative, unbiased in signal absence, and minimizing both the variance due to noise alone and the bias due to a symmetrically observed signal. This estimator depends on the signal's normalized envelope but not on the amplitude. The derived estimators are used in determining structures of two noncoherent adaptive signal detectors in unknown level additive noise. For independent noise samples, the first detector reduces to a uniform integrator with cell-averaging CFAR circuit, and the second reduces to a filter matched to the squared signal envelope but with quasi NAR AGC which can regulate faster noise power fluctuations. © 1990 IEEE