Bayesian signal detection for multiple aspect targets with an uncertain look angle

An optimal signal detection theory (likelihood ratio) approach is presented for the determination of the presence or absence of a target observed at multiple aspects in noise, where there is uncertainty in the initial look angle at which the aspects are observed. Potential targets may be interrogated at any number of aspect angles, and receiver operating characteristics (ROC's) are presented as a function of the number of aspects observed. In order to obtain the effect on performance of the number of aspect angles and other characteristics of the signal, ROC comparisons are made for the same total signal-to-noise ratio (SNR), rather than the average SNR per aspect. Target returns from a real multiple aspect target data set, consisting of acoustic backscatter returns from several objects suspended in a tank of water, are utilized in a detection simulation. The results using the real data indicate that, for the same total signal-to-noise ratio, detection does not necessarily improve with an increasing number of look angles. Theoretical analysis shows that optimum detection for this situation occurs when the signals consisting of multiple aspects, but at different initial look angles, are highly correlated. This conclusion is supported by the real data and shows that detection performance does not necessarily improve with the number of multiple aspect angles observed, for a given total signal-to-noise ratio, when the initial look angle is uncertain.

Duke Scholars

Author Stacy L. Tantum Electrical and Computer Engineering

Author Loren W. Nolte Electrical and Computer Engineering