摘要: A systematic approach is developed as a means of characterizing optimal and conventional space-time detection of a Gaussian signal in colored Gaussian flow noise with line arrays. Discrete arrays of sensors, as well as continuous ... 展开 A systematic approach is developed as a means of characterizing optimal and conventional space-time detection of a Gaussian signal in colored Gaussian flow noise with line arrays. Discrete arrays of sensors, as well as continuous finite observation intervals, are considered. Using the likelihood ratio as a starting point, exact solutions are formulated, specifying the optimal instrumentation for a rational noise model. These solutions provide the theoretical framework to determine upperbound detection performance and permit examination of the role played by the noise spectra in influencing the optimal instrumentation and its performance. The study describes potential gains achievable with optimal processing over conventional processing. Flow noise spectral parameters are identified, for which either the detection gains with optimal processing are significant or conventional processing is sufficiently close to optimal. The best narrowband robust performance is nearly achieved with conventional processing alone, especially if the spatial bandwidth-length product is large. A simple modification is examined of basic Eckart frequency filtering at a conventional beamformer output. This modification recovers a substantial part of the detection loss incurred with basic Eckart filtering and conventional beamforming. Wideband performance with conventional beamforming followed by modified Eckart filtering remains essentially invariant as the noise model ranges over a large class of rational spectra, including a simple Butterworth spectrum at one extreme and a Gaussian spectrum at the other. 收起