Adaptive detection for unknown noise power spectral densities

The detection of a known broadband signal in colored noise of unknown power spectral density is addressed. Motivated by the consistency of the integrated periodogram, a new detector is proposed. Its asymptotic performance is proven to be only slightly poorer than the optimal but unrealizable Neyman-Pearson detector. It also possesses the CFAR property asymptotically and should therefore be quite valuable in practice. For finite data records, it is shown by computer simulation to significantly outperform the conventional matched filter (without prewhitening) under realistic conditions encountered in practice.