Acoustic scene modeling for echolocation in bottlenose dolphin

Abstract

The biosonar of bottlenose dolphins has been shown to have excellent target discrimination performance in cluttered environments, but how the animals process the flow of information used in their adaptive searching process is still an open question. In this work, we present a physics-based model of the process echolocating dolphins may use during target discrimination tasks and validate the model with experimental results. Assuming that dolphins emit biosonar clicks and continuously compare the return echoes with learned representations of echoes from familiar objects, we define a likelihood parameter, a metric that quantifies this comparison. We show that the dolphins’ adaptive search behavior during echolocation can be interpreted as an effort to maximize the likelihood parameter. Our model is validated experimentally using target discrimination tasks in which a dolphin correctly/incorrectly located a given target in the presence of distraction objects in an open lagoon while it was wearing eyecups. The results show that the dolphin positioned itself where the likelihood parameter was maximized demonstrating that our model explains well the dolphin’s behavior during the tasks. [Work supported by the Department of the Navy Grant No. N00014-18-1-2069.]

Metamaterial characterization from far-field acoustic wave measurements using convolutional neural network

“…However, near-field measurements are not always available, e.g., in-vivo tissue diagnostics. Nevertheless, it has also been shown that echolocating animals use far-field scattered wave to discriminate acoustically small objects such as fish [17], which suggests that the information contained in the diffracted fields could be extracted from far-field single point measurements.…”