Asymptotically optimum detector of an unknown sinusoid in AWGN

A general approach is proposed to resolve the wideband detection problem, which utilizes data in the correlation domain via autoregressive modeling. The structures of Gaussian autoregressive processes are reviewed and applied to the modeling of a sinusoid in additive white Gaussian noise (AWGN). Based upon the output sequences of this adopted model, optimal hypothesis-testing tools are employed, leading to a novel scheme, namely the multiple-correlation-coefficient detector. For a properly selected model, this statistic is shown to be competitive to the spectral-maximum detector. This fact is established analytically as well as through extensive simulations. Connections to the other detectors in the correlation domain are also established by means of this model-based approach