ORCA-SPY enables killer whale sound source simulation, detection, classification and localization using an integrated...

Area of Science:

• Marine bioacoustics and bioacoustics software development.
• Computational bioacoustics and signal processing.
• Wildlife monitoring and conservation technology.

Background:

• Acoustic identification of individual animals is crucial for understanding communication but challenging underwater.
• Lack of ground truth localization data hinders evaluation of passive acoustic monitoring (PAM) methods for marine species.
• Killer whale (Orcinus orca) vocalizations are complex, requiring advanced techniques for individual identification and localization.

Purpose of the Study:

• To present ORCA-SPY, a new automated framework for simulating, classifying, and localizing killer whale sounds.
• To generate realistic, ground truth localization data for evaluating killer whale acoustic monitoring systems.
• To provide an open-source tool adaptable for various recording conditions and marine species.

Main Methods:

• Developed ORCA-SPY, a framework integrated into PAMGuard, featuring automated sound source simulation and classification.
• Employed a hybrid approach combining ANIMAL-SPOT (a deep learning orca detector) with Time-Difference-Of-Arrival (TDOA) localization.
• Evaluated the system using simulated multichannel audio streams and field tests in diverse underwater environments.

Main Results:

• Achieved a 94.0% detection rate with an average localization error of 7.01 meters on simulated data across various conditions.
• Field tests demonstrated practical performance with average localization errors of 29.19m (lab) and 20.01m (expedition).
• The framework proved effective in real-world deployments, showing median errors as low as 11.01 meters.

Conclusions:

• ORCA-SPY successfully simulates and localizes killer whale vocalizations with high accuracy, addressing the challenge of limited ground truth data.
• The open-source framework offers a valuable tool for researchers studying marine mammal acoustics and behavior.
• ORCA-SPY's adaptability suggests potential for broader applications in passive acoustic monitoring of other species.