Information transfer of broadband sonar echoes

Abstract

The ability to compute information transfer of a broadband signal corrupted with additive noise is straightforward, but the case of signal corrupted by convolution with a stochastic, propagation impulse-response is more complicated. The known, integral solution is presented along with numerical methods for estimating differential entropy and mutual information using kernel density estimators. Numerical examples using simulations of echo and propagation responses are shown. Results are shown using echo data from Clutter09, an ocean experiment performed with an echo repeater. These echoes were generated with a known, modeled target response. The mutual information between the modeled target response and the echoes, which propagated through ocean environments, is also shown. These results are discussed with respect to developing a method of obtaining an information sonar equation that could be used to estimate the sonar parameters required to perform an information based task, such as signal classification. [Research funded by the Office of Naval Research.]