Threshold Region Performance Prediction for Adaptive Matched Field Processing Localization

The full-field modeling of matched field processing (MFP) provides potentially large gains over traditional beamformers in passive sonar signal processing. MFP localization gains are manifested by superior estimation of source range and novel estimation of source depth. However, MFP localization is limited in practice by high spatial ambiguities in the MFP output. These ambiguities can result in global localization errors at low signal-to-noise ratios, errors which must be accounted for to predict MFP localization performance accurately. This work uses the so-called method of interval errors (MIE) to predict MFP mean-squared error localization performance well into the “threshold region” where global errors dominate. New results enable highly accurate threshold region MIE predictions for both conventional beamforming and Capon-MVDR adaptive beamforming in the simple signal-in-white-noise case. A variation of the standard MIE technique is shown to produce equally accurate performance predictions when both mismatch and colored noise are introduced.