General principles of Matched Field Processing are explained, including model and processor selections, search domains, effects of grid spacing and environmental mismatch. Matched Mode analysis of the 93 MFP Workshop and Hudson Canyon datasets is then performed using four inversion techniques: direct inversion, eigenvalue, singular value decomposition and neutral regression. Strengths and weaknesses of each method are outlined. Comparisons are performed using ambiguity surfaces and modal coefficient graphs for each inversion method, at each of the eight frequencies of the Hudson Canyon dataset. |