Depth classification based solely on incoherent sonar information

Abstract

Most work on sonar contact depth estimation has been based on deterministic, coherent propagation modeling of the sound channel, e.g., matched-field processing. This has met with limited success due to the inability to precisely predict the sound-pressure field in realistic scenarios. This paper addresses the problem of using probabilistic, incoherent information from the sonar itself for depth classification with active sonar, without having to depend on precise and accurate propagation models and ancillary environmental measurements. In particular, the problem of deciding whether or not a given contact is on the bottom based solely on the sonar data is looked at, i.e., without resorting to the use of any additional environmental measurements or predictive models. To do this, the in situ local bottom reverberation is used to calibrate the channel. Probability theory is explored to provide a theoretical basis for the development of a single-hypothesis decision metric to best exploit this information. The methods are tested on a combination of broadband sonar data and detailed ocean simulations. The particular metric proposed for this problem seems to be important for achieving good performance, and may be of some interest in its own right for other types of single-hypothesis decision problems.