A Matched Subspace Approach to Depth Discrimination in a Shallow Water Waveguide

An application of the Scharf-Friedlander matched subspace detector [1] to the problem of trapped mode subspace discrimination in shallow water acoustics is investigated. The "trapping" of low order modes below the thermocline in a downward refracting shallow channel has been shown to be useful to a class of physics-based sonar discrimination algorithms [2]. It will be shown that by applying a subspace orthogonalization step to a suitably rank-reduced normal mode subspace model, an optimum "trapped" subspace discrimination algorithm is possible even when the trapped and non-trapped subspaces exhibit significant overlap. Data results illustrating performance improvement over a previous sub-optimum LRT implementation that ignored subspace overlap for a shallow water, thinned-VLA case will be shown.