Threshold Region Performance Prediction for Adaptive Matched Field Processing Localization

摘要

Matched field processing (MFP) provides a means of attaining the full gains available from the shallow-water acoustic channel in passive sonar signal processing. By modeling the full field structure of acoustic signals propagating in the ocean MFP offers the potential for both detection gain (through its better signal model) and localization gain (through its additional discrimination capability in range and depth) over traditional planewave processing. However, high spatial ambiguities and mismatch present formidable challenges in practice limiting the performance gains that are realistically achievable with MFP. Prediction of MFP localization performance is a challenging problem. MFP replica (steering) vectors can be highly ambiguous in range and depth resulting in significant non-local estimation errors at low signal-to-noise ratios (SNRs)-errors not modeled by traditional localization measures such as the Cramer-Rao bound. Recent work has demonstrated the accuracy of an interval-error-based method referred to herein as the 'method of interval errors' (MIE), in predicting mean- squared error localization performance well into the threshold region where non-local errors may dominate. This work uses the MIE to predict the mean-squared error accuracy of MFP range and depth estimates for two well-known approaches: (i) conventional beamforming (equivalent to maximum likelihood estimation for white noise) and (ii) Capon- MVDR adaptive beamforming. Simulation results will characterize localization performance as a function of SNR, for apertures and environments of interest. Particular attention will be given to the 'threshold SNR' (below which localization performance degrades rapidly due to global estimation errors) and to the minimum SNR required to achieve acceptable range/depth localization. Initial work will also be presented assessing the MIE's potential to characterize localization performance in the presence of mismatch.