Acoustic Source Drone Detection System Using Tetrahedral Microphone Array and Deep Neural Networks

Area of Science:

• Aerospace Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Unmanned Aerial Vehicles (UAVs) integration poses security challenges.
• Conventional drone detection methods (radar, optical) have limitations.
• Acoustic analysis offers a passive detection method but faces localization challenges.

Purpose of the Study:

• To develop a deep learning framework for accurate 3D localization of UAVs.
• To address limitations of acoustic localization with directional sensors.
• To improve security for critical infrastructure and privacy against UAV threats.

Main Methods:

• A deep learning framework fusing raw acoustic data with sensor geometry metadata.
• A neural network architecture designed for acoustic signal processing.
• A composite loss function optimizing planar and altitude coordinates.

Main Results:

• The proposed system achieves robust localization performance in complex acoustic environments.
• Effective mitigation of spatial irregularities from ad hoc sensor deployment.
• Enhanced precision in three-dimensional localization of drones.

Conclusions:

• Deep learning offers a viable solution for accurate UAV acoustic localization.
• The framework successfully overcomes challenges posed by directional sensors and complex environments.
• This technology enhances airspace security and privacy protection against UAVs.