Autoencoder-Based Signal Modulation and Demodulation Methods for Sonobuoy Signal Transmission and Reception

Area of Science:

• Underwater acoustics
• Signal processing
• Machine learning

Background:

• Conventional sonobuoys transmit acoustic data using modulation techniques like FSK, leading to high data volume and low security.
• Existing methods face challenges with large information transmission and limited security due to simple modulation/demodulation.

Purpose of the Study:

• To propose an autoencoder-based method for encoding sonobuoy signals into low-dimensional latent vectors.
• To enhance signal security and reduce transmission data volume significantly.
• To introduce a denoising autoencoder to mitigate ambient noise in reconstructed signals.

Main Methods:

• Utilized autoencoders to encode underwater acoustic signals into low-dimensional latent vectors for transmission.
• Employed autoencoders for decoding latent vectors, improving signal security and data compression.
• Simulated bistatic active and passive sonobuoy environments to evaluate performance.
• Developed a denoising autoencoder to reduce ambient noise in reconstructed signals.

Main Results:

• The proposed autoencoder successfully restored original signals from low-dimensional latent vectors with approximately 4% error.
• Data transmission volume was reduced by approximately a factor of one hundred compared to conventional methods.
• The denoising autoencoder effectively reduced ambient noise, validated through spectrogram analysis and SNR measurements.

Conclusions:

• Autoencoder-based signal processing offers a secure and efficient method for sonobuoy data transmission.
• The denoising autoencoder enhances the quality of reconstructed signals by reducing ambient noise.
• This approach significantly improves the performance and capabilities of underwater acoustic surveillance systems.