Cyclostationary-Signal Detection and Parameter Estimation

Several algorithms for the estimation of the multipath-channel parameters by noisy input/output measurements have been proposed in [J2]–[J4], [BC1]. The proposed methods, that exhibit different time-delay resolution, operative range, and disturbance tolerance, exploit the cyclostationarity properties exhibited by virtually all man-made signals to perform estimates intrinsically immune to the effect of noise and interference, provided that a cycle frequency exists that is not shared by the disturbance and the useful signal. Such an assumption is relaxed in [J4], where the time-difference-of-arrival (TDOA) of a signal impinging on two sensors is performed provided that it exists at least a portion of the useful signal cyclic spectrum that is not overlapped with the disturbance cyclic spectra.

In [C4], the spectral correlation property is exploited to detect cyclostationary signals in white non-Gaussian noise of unknown level. In [J1], a generalized log-likelihood ratio test (GLLRT) is derived for the problem of the detection of a known signal embedded in cyclostationary white generalized Gaussian noise.