C. Feuillade
Naval Res. Lab., Stennis Space Center, MS 39529-5004
The principal application of matched-field processing (MFP) over the last several years has been the determination of acoustic source information, primarily location parameters. In shallow water, where the channel characteristically promotes strong boundary interactions, the successful implementation of MFP methods requires accurate environmental information to accurately model the acoustic field. When the environmental data used to calculate the replica field are incomplete or inaccurate (leading to data ``mismatch'') MFP performance can be seriously impaired. While it is clear that errors in all of the contingent acoustic parameters will affect MFP, much work has been devoted to establishing a scale of relative importance for these quantities, so that appropriate effort can be expended to obtain reliable values for the most important ones and to determine the effects that inaccuracies in them will have. In this paper, a review is presented of some of the numerical and experimental studies that have been performed to understand the data mismatch issue; and also a discussion of how this has motivated and directed the search for ``robust'' MFP methods that provide stable performance even when the environment is not known well will be presented. One particular class of estimators, involving ``dimensionality reduction'' eigenvector techniques, attempt to overcome this problem by isolating the effects of mismatch in the vector space outside of that which is processed. This talk describes how this approach may be adapted to do the inverse problem: to use MFP to provide data about the acoustic environment. [This work was supported by the Naval Research Laboratory.]