Adaptive sonar detection performance when the signal wavefront and noise covariance are uncertain

This paper considers adaptive sonar array detection performance when both the multipath ocean environment and the noise field directionality are uncertain. Recent analytic results for adaptive matched-subspace detector (AMSD) performance with multi-rank signals in Gaussian noise of unknown covariance are applied to the passive sonar problem. In particular, a methodology for comparing analytically predicted detection performance to that achieved using a limited amount of real horizontal acoustic array data collected off San Diego is presented. Detection performance results with real data are shown to be in good agreement with theory provided that sufficient signal-subspace rank is assumed and that the signal does not contaminate the assumed signal-free noise training data. Discrepancies between theory and data are largest when the SNR was high and the number of snapshots available to estimate the noise covariance matrix was low.