Abstract
Effective underwater acoustic communications requires source symbol decisions in the presence of an uncertain space and time varying acoustic response function. A hierarchical mixture Gaussian model is useful for modeling both the sparsity of arrivals as well as their spread in angle, Doppler and propagation delay [Canadian Acoust. 40]. In this framework, the delay spread, the degree of sparsity, and the Doppler spread all must be marginalized based on the observed data. We discuss the degree of sparsity as well as the correlation among multi-path arrival times and how these uncertain features in the response function can be efficiently treated in a computationally reasonable and statistically efficient manner via the hierarchy. The approach relies on iterative coherent symbol decisions with an empirical Bayes approach to estimating the hyper-parameters of the model permiting flexibility to adapt to environmental conditions. It is shown that coherent multi-path combining and Doppler compensation are possible at extremely low signal to noise ratios (i.e., < −18 dB), at ranges in excess of 1 km and with throughputs exceeding 100 bps with single element reception. Results are shown for large bandwidth M-ary orthogonal sequences tailored only to a maximum allowable multipath spread.