George E. Ioup
Univ. of New Orleans, Dept. of Phys., New Orleans, LA 70148
Lisa A. Pflug
Naval Res. Lab. Detachment--Stennis Space Center
Juliette W. Ioup
Univ. of New Orleans, New Orleans, LA 70148
Robert L. Field
Naval Res. Lab. Detachment--Stennis Space Center
Narrow-band processing is commonly applied to achieve signal-to-noise ratio (SNR) advantage when the signal is concentrated in frequency and the noise bandwidth is broader. The matched filter (ordinary cross correlator) rejects noise outside the signal passband when the source is known. If only the passband of an energy transient is known, but other signal details are unknown, prefiltering of the data can sometimes improve ordinary correlation detectors and in some cases, produce even greater improvement in higher-order correlation detectors. For example, in the cross power spectrum, if the passband of the signal is half the passband of the noise and the noise is not preferentially concentrated in frequency, for a given probability of detection and fixed probability of false alarm, there will be a potential two-to-one maximum improvement in SNR with prefiltering. With one-dimensional prefiltering, for the cross bispectrum the potential maximum improvement rises to four-to-one and for the trispectrum, to eight-to-one. Larger maximum improvements are possible with higher-order prefiltering. In practice, one expects improvements less than the maximum. The prefiltering makes possible substantial improvements in detector performance which are shown using Monte Carlo simulations. Improvements of as much as a factor of 4 in SNR over the unfiltered case have been observed. [Work supported by ONR/NRL-SSC.]