Drones Detection Using a Fusion of RF and Acoustic Features and Deep Neural Networks.

Area of Science:

• Electrical Engineering
• Computer Science
• Aerospace Engineering

Background:

• Drones (Unmanned Aerial Vehicles) are increasingly used in various sectors, but their misuse poses significant security risks.
• Effective remote detection of potentially threatening drones is crucial for safety and security.
• Current detection methods may struggle in complex environments or with low signal-to-noise ratios (SNR).

Purpose of the Study:

• To develop a novel, automatic drone detection system.
• To enhance drone classification accuracy by fusing radio frequency (RF) and acoustic signals.
• To evaluate the performance of classical and deep machine learning techniques for drone detection.

Main Methods:

• Utilized both radio frequency (RF) communication signals and acoustic signals from UAV rotor sounds.
• Applied classical (Support Vector Machine - SVM) and deep learning (Convolutional Neural Network - CNN, Recurrent Neural Network - RNN, specifically LSTM) classifiers.
• Fused RF and acoustic features for improved drone classification.
• Evaluated the approach using common drone datasets with varying SNR levels.

Main Results:

• The proposed fused-feature approach demonstrated superior accuracy compared to existing methods.
• Achieved approximately 91% classification accuracy using an LSTM network at a low SNR of -10 dB.
• The system proved effective, particularly in low SNR scenarios, highlighting robustness.

Conclusions:

• The fusion of RF and acoustic signals with advanced machine learning offers an effective solution for automatic drone detection.
• The developed method shows significant promise for enhancing drone security and mitigating potential threats.
• The approach is particularly valuable for reliable drone detection in challenging environments with weak signals.