Computer models for masked hearing experiments with beluga whales (Delphinapterus leucas)

Area of Science:

• Marine biology
• Bioacoustics
• Computational acoustics

Background:

• Environmental assessments require understanding manmade noise impacts on marine mammal vocalizations.
• Direct animal experiments are time-consuming, costly, and often impractical for such assessments.

Purpose of the Study:

• To evaluate the efficacy of software models in predicting noise interference with animal vocalizations.
• To compare model performance against empirical data from beluga whale hearing experiments.

Main Methods:

• Utilized signal processing techniques including matched filter, spectrogram cross-correlation, and critical band cross-correlation.
• Employed a back-propagation neural network to detect beluga vocalizations in various ocean noise conditions.
• Compared model outputs with masked hearing experiments conducted on a beluga whale.

Main Results:

• The back-propagation neural network demonstrated the highest fidelity in simulating the beluga whale's masked hearing data.
• All tested models showed capability in detecting beluga vocalizations amidst different ocean noise types.
• The neural network approach provides a promising method for predicting noise-vocalization interference.

Conclusions:

• Software models, particularly artificial neural networks, can reliably estimate the effects of anthropogenic noise on marine mammal communication.
• This approach offers a feasible and efficient alternative to traditional animal testing for environmental noise impact studies.
• Findings support the use of computational models in marine mammal conservation and acoustic management strategies.