Understanding auditory distance estimation by humpback whales: a computational approach.

Area of Science:

• Marine bioacoustics
• Animal behavior
• Signal processing

Background:

• Sound ranging is crucial for marine mammals, relying on predictable sound attenuation patterns.
• Humpback whale songs are complex vocalizations used for communication over long distances.

Purpose of the Study:

• To investigate if humpback whales use frequency degradation cues for ranging singing whales in coastal waters.
• To assess the effectiveness of neural networks in classifying sound distances based on frequency content.

Main Methods:

• Measuring long-range sound propagation in coastal waters.
• Training multi-layer and single-layer perceptron neural networks to classify sounds by distance.
• Processing recordings with a computational model of the humpback whale's auditory system.

Main Results:

• A multi-layer neural network successfully classified sound distances using frequency content, confirming its utility for ranging.
• Normalizing sounds by ambient noise improved distance estimation accuracy for single-layer perceptrons.
• Physiologically based auditory models, despite reduced information, led to more accurate distance estimations by neural networks.

Conclusions:

• Underwater sound propagation effects on frequency content provide sufficient cues for humpback whales to estimate source distance.
• Whale auditory system processing may enhance the ability to perceive distance-dependent frequency changes in songs.
• Humpback whales likely utilize frequency degradation as a key mechanism for ranging singing conspecifics.