Sparse acoustic response function estimation with a mixture Gaussian model

Abstract

Underwater acoustic response functions at high frequencies and large bandwidths exhibit significant spatio-temporal variability that depends greatly on volume, boundaries, and the source–receiver motion. Considered here is a mixture Gaussian assignment over Doppler, beam-angle, and channel bandwidth employed to describe the behavior of the sparse acoustic response function over received signal duration, aperture, and bandwidth. Accurate modeling of the dependence between the mixture components can be handled by considering dependence among neighboring indicator variables of the mixture assignment. This allows for a more accurate and adaptable description of the natural persistence that acoustic paths exhibit and improve channel estimation quality. This adaptive structure is applied to underwater M-ary spread spectrum acoustic communication transmissions during the MACE10 experiment of the coast of Martha's Vineyard at near 10 kHz of bandwidth and at ranges of 1 and 2 km. Posterior conditional expectations of the acoustic response are compared with least squares type estimates and performance is quantified in terms of the observed bit error rate (BER) as a function of received SNR. A BER = E-6 at rSNR = −16 dB for 12 element combining is demonstrated. [This work is supported by the Office of Naval Research and by NISE BAR.]