ACOUSTIC IMAGING OF RANDOMLY-DISTRIBUTED UNDERWATER SOURCES

Abstract

Snapping shrimp noise is thought to be a major component of ambient noise at high frequencies (2 kHz - 300 kHz) in warm shallow water. We have designed an experiment to investigate its temporal and spatial distribution and variability. This information has application to many underwater acoustic systems in providing background knowledge about the structure of this class of noise. The experiment requires a stochastic tomographic inversion to estimate the source structure from data collected by several directional hydrophones. Existing tomographic techniques have been extended to benefit from the broadband nature of the data. The experiment has been modelled by including the receiver directionality, possible source distributions and several sources of noise in the signal processing. Source distribution models have been developed and calibrated for consistency with actual shallow-water acoustic records. We develop three noise models: electronic noise, beam-pattern sensitivity noise and hydrophone direction uncertainty. The simulation results indicate that our stochastic inversion algorithm is robust to reasonable levels of these types of noise. We anticipate that the experiment will be able to image shrimp noise intensity on the seabed over an area of some 350m2 with a resolution of 3.5m2 and an rms error below 20% (approximately 1.8 dB).