A source separation approach to enhancing marine mammal vocalizations

Area of Science:

• Marine bioacoustics
• Signal processing
• Underwater acoustics

Background:

• Passive acoustic monitoring (PAM) of marine mammals is crucial but often hampered by noise interference, such as from vessels.
• Conventional noise suppression techniques like beamforming require complex arrays and may struggle with low signal-to-noise ratios (SNR).
• Developing effective underwater signal enhancement methods is essential for accurate marine mammal population assessments.

Purpose of the Study:

• To propose and implement an alternative multi-channel underwater signal enhancement technique.
• To develop a blind source separation (BSS) algorithm for extracting marine mammal vocalizations from noisy two-channel measurements.
• To assess the algorithm's effectiveness in improving the detection range of passive acoustic detectors.

Main Methods:

• A novel blind source separation algorithm utilizing a robust decorrelation criterion was derived and implemented.
• The algorithm was adapted to an unsupervised framework by incorporating the supervised affine projection filter update rule to handle temporally correlated data.
• The method was evaluated using real West Indian manatee (Trichechus manatus latirostris) vocalizations and recorded watercraft noise.

Main Results:

• The proposed BSS algorithm successfully separated manatee vocalizations from background noise, even in low SNR conditions.
• The algorithm demonstrated suitability for low SNR measurements, a common challenge in marine acoustic monitoring.
• The method improved the detection range of a passive acoustic detector by an average factor of five for input SNRs between -10 and 5 dB.

Conclusions:

• The developed blind source separation algorithm offers a robust and effective alternative for underwater acoustic signal enhancement.
• The algorithm's ability to improve detection range significantly enhances the utility of passive acoustic monitoring systems, especially with limited hydrophone arrays.
• This approach is particularly valuable for monitoring species like the West Indian manatee in noisy environments, potentially increasing conservation efforts.