Presentation Details
| Assessing the distinctiveness of synthetic vocal identity signals produced by stochastic generation Grethel A Juarez1, Raneem Samman1, 2, Brittany Coppinger1, 3, Grace Smith-Vidaurre1-3. 1Department of Integrative Biology, Michigan State University, East Lansing, MI, USA.2Department of Computational Mathematics, Science, and Engineering, East Lansing, MI, USA.3Ecology, Evolution, and Behavior Program, Michigan State University, East Lansing, MI, USA |
Abstract
Animals can use vocalizations to communicate identity for social recognition. Computational tools can be used to yield insight into the diverse functions of vocalizations for identity information transmission. However, it is difficult to validate detection of vocal identity information using empirical datasets. While synthetic vocalizations can aid in validation, information detection may vary depending on the generative process used to create synthetic signals. We used a stochastic generation process with the package paRsynth to create character string datasets representing synthetic vocal identity signals. We tested how the distinctiveness of group identity information was impacted by parameters used to encode information. We performed simulations to test how the number of unique symbols in strings, string length, and the size of signal sets, impacted the information capacity of the resulting synthetic datasets. By applying information theoretic tools, we found that small signal sets of long strings with more unique symbols were most distinctive. These findings provide a baseline for future work using agent-based experiments to generate vocal variation.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
Content Locked. Log into a registered attendee account to access this presentation.