File(s) not publicly available

Biomimetic detection of dynamic signatures in foliage echoes

journal contribution
posted on 06.10.2021, 00:29 by Ananya Bhardwaj, Mohammad KhyamMohammad Khyam, Rolf Müller
Certain bat species (family Rhinolophidae) dynamically deform their emission baffles (noseleaves) and reception baffles (pinnae) during echolocation. Prior research using numerical models, laboratory characterizations, and experiments with simple targets have suggested that this dynamics may manifest itself in time-variant echo signatures. Since the pronounced random nature of echoes from natural targets such as foliage has not been reflected in these experiments, we have collected a large number (>55,000) of foliage echoes outdoors with a sonar head that mimics the dynamic periphery in bats. The echo data was processed with a custom auditory processing model to create spike-based echo representations. Deep-learning classifiers were able to estimate the dynamic state of the periphery, i.e., static or dynamic, based on single echoes with accuracies of up to 80%. This suggests that the effects of the peripheral dynamics are present in the bat brains and could hence be used by the animals. The best classification performances were obtained for data that was obtained within a spatially confined area. Hence, if the bat brains suffer from the same generalization issues, they would have to have a way to adapt their neural echo processing to such local fluctuations to exploit the dynamic effects successfully.

History

Volume

16

Issue

4

Start Page

1

End Page

15

Number of Pages

15

eISSN

1748-3190

ISSN

1748-3182

Location

England

Publisher

IOP Publishing

Language

eng

Peer Reviewed

Yes

Open Access

No

Acceptance Date

16/04/2021

External Author Affiliations

Virginia Tech, USA

Era Eligible

Yes

Medium

Electronic

Journal

Bioinspiration and Biomimetics

Article Number

046026