Machine Learning Based Biomimetic Underwater Covert Acoustic Communication Method Using Dolphin Whistle Contours.

Area of Science:

• Bioacoustics
• Underwater Acoustic Communications
• Machine Learning

Background:

• Biomimetic covert communication methods using dolphin whistles currently suffer from low mimicry (DoM) due to signal pattern differences.
• Existing methods struggle to balance high DoM with low bit error rates (BER) in underwater acoustic environments.

Purpose of the Study:

• To propose a novel biomimetic communication method that enhances the degree of mimic (DoM) while maintaining a low bit error rate (BER).
• To improve the performance of underwater acoustic covert communication systems by leveraging dolphin whistle characteristics.

Main Methods:

• Utilized diverse contours of real dolphin whistles and link information between consecutive whistles for signal transmission.
• Employed machine learning-based whistle detectors, incorporating inter-whistle link information, for signal reception.
• Validated the proposed method through computer simulations and practical ocean experiments.

Main Results:

• The proposed method demonstrated significantly better bit error rate (BER) performance compared to conventional approaches.
• Ocean experiments recorded a BER of 0.002 for the proposed method, a substantial improvement over the conventional Deep Neural Network (DNN) detector's BER of 0.36.

Conclusions:

• The novel biomimetic communication method effectively preserves a high degree of mimicry (DoM) with a low bit error rate (BER).
• The integration of whistle contours and link information, coupled with machine learning detectors, offers a superior solution for underwater acoustic covert communications.