Autocorrelation based denoising of manatee vocalizations using the undecimated discrete wavelet transform

Area of Science:

• Marine Biology
• Acoustics
• Signal Processing

Background:

• Watercraft collisions pose a significant threat to West Indian manatee populations.
• Passive acoustic monitoring using manatee vocalizations is a proposed method for collision avoidance.
• Existing denoising techniques struggle with the low signal-to-noise ratio (SNR) typical of manatee vocalizations.

Purpose of the Study:

• To develop an effective denoising algorithm for West Indian manatee vocalizations.
• To improve the performance of passive acoustic detection systems for manatee conservation.

Main Methods:

• A novel wavelet-based denoising scheme was developed, leveraging the autocorrelation function of manatee vocalizations.
• The method combines the strengths of nonlinear wavelet transform denoising with signal-specific characteristics.
• Performance was evaluated against traditional linear filtering methods.

Main Results:

• The proposed wavelet-based algorithm significantly outperforms linear filtering in denoising manatee vocalizations.
• The method is effective even at signal-to-noise ratios below 0 dB.
• Improved vocalization detection range was achieved.

Conclusions:

• The developed denoising scheme offers a viable solution for enhancing passive acoustic monitoring of West Indian manatees.
• This advancement can contribute to reducing manatee mortality by improving early warning systems.
• Exploiting signal-specific properties like autocorrelation is crucial for effective bioacoustic signal processing.