Effect of Environmental Variability on Model-Based Signal Processing: Review of Experimental Results in the Mediterranean

Abstract

Over the last decade experiments were conducted in the Mediterranean for localising controlled sound sources and for deducing bottom geo-acoustic properties from the measurement of the acoustic-channel impulse response over a broad frequency range. Inversion of the large time-bandwidth-product signals transmitted to probe the medium was performed using a coherent receiver, the model-based matched filter. The reference channels incorporate Green’s function models for partially known or hypothesised environmental conditions and source parameters (range, depth and Doppler). The search terminated when most of the time-spread energy on single or multiple elements of the receive array was recombined coherently into a single peak. Although performance limitations were imposed by environmental and modelling uncertainties the model-based processor was always seen as intrinsically robust to the acoustic-signal variability encountered in our experiments. Substantial processing gains were obtained in most situations depending on the time dispersion, spatial diversity, predictability and coherence of the specific acoustic channel. Most importantly, correct and stable inversion results were possible even from a sparse but representative set of hydrologic data.